Fluorescence imaging analysis of depth‐dependent degradation in photovoltaic laminates: insights to the failure

Accurate evaluation of the reliability of photovoltaic (PV) packaging materials is critically important for the long‐term safe operation of modules. However, the complexity of the laminated systems due to their multilayered and multicomponent structures and diverse aging mechanisms makes a thorough system evaluation very challenging, especially when the degradation is non‐uniform through the thickness. In such a case, neither surface nor bulk measurements can present a clear picture of the degradation profile. In this study, fluorescence imaging was developed to visualize the degradation depth‐profiles of an aged laminated PV system. A glass/ethylene vinyl acetate (EVA) encapsulant/poly(ethylene terephthalate) (PET)‐PET‐EVA (PPE) backsheet laminate was weathered with the glass‐side facing an ultraviolet (UV) light source for 3840 h. Cross‐sectional fluorescence images revealed a non‐uniform distribution of degradation species across the thickness of the EVA encapsulant, providing greater insight into the mechanisms of degradation, which are unavailable by traditional bulk‐based methods. In addition, strong fluorescence emissions were observed from the two thin adhesive layers of the aged backsheet, indicating severe degradation of the adhesives and a potential for interlayer delamination. This method is further confirmed with other microscale characterization techniques. The changes in optical (yellowness index), chemical (oxidation, UV absorber concentration), mechanical (Derjaguin‐Muller‐Toporov modulus), and thermal (melting enthalpy) properties of the EVA encapsulant were found to be related to fluorescence profiles, following the attenuation of UV light. This study highlights that fluorescence imaging is a spatially‐resolved and sensitive method for rapid failure assessment and in‐depth mechanism study for complex PV‐laminated system.


| INTRODUCTION
To ensure the long-term reliability of photovoltaic (PV) modules, a multilayer encapsulant system is used to create multiple physical barriers for the solar cell to prevent damage from UV light, moisture or oxygen. 1 However, the polymeric materials used for encapsulants and backsheets are susceptible to degradation and their degradation or failures (such as discoloration, cracking or delamination) tend to initiate at their surfaces or weak interfaces. [2][3][4][5] Gradual progression of such localized degradation to the whole packaging system would lead to non-uniform failure, 6 which finally induces dramatic loss in the module power efficiency and the catastrophic failure of the whole system. Despite the great necessity to evaluate the inhomogeneous degradation behavior to predict the reliability and lifetime of PV systems, most studies are mainly confined to bulk property changes. [7][8][9][10][11] Investigations on the inhomogeneity of degradation within the layers of aged packaging materials for PV application is still scarce.
Gradient degradation of polymers, induced by the ultraviolet (UV) light attenuation or/and the diffusion limited effects of oxygen and moisture exposure, is ubiquitous and has been studied extensively in past decades. [12][13][14][15] Nevertheless, the complex construction of PV modules, with multiple layers and components, makes it especially challenging to quantify the structure and property degradation of individual components/layers during their service lifetime, and to correlate them with the likelihood of failure. Several spectroscopic imaging techniques, such as Fourier-transform infrared (FTIR) spectroscopy and Raman spectroscopy have been used to examine the degradation gradient of multilayer PV laminates. [16][17][18] While these techniques are capable of mapping degradation, their limited spatial resolution, low signal-to-noise ratios, or time-consuming measurement procedures have hampered wider application of these techniques. For example, although Raman spectroscopy has superior lateral resolution compared to FTIR spectroscopy, significantly enhanced fluorescence background emitted from aged materials dominates Raman spectra and obscures spectroscopic information. 18 While fluorescence emission hinders the utility of Raman spectroscopy, it can provide valuable information on evaluating polymer degradation, 19,20 as it is sensitive to conjugated products which often form during degradation. Approaches using fluorescence emission are promising to determine the degree of degradation in an EVA encapsulant 20,21 and identify cell cracks 22 in PV modules. For example, Röder et al. [23][24][25] successfully used fluorescence spectroscopy with a lateral resolution of about 1 mm to characterize the degradation of EVA from the edge to the center of PV modules after outdoor and accelerated aging (damp-heat and UV aging). Although their work did not provide depth-dependent information, it has laid groundwork for using fluorescence measurement as a tool for characterization of degradation of PV polymers.
In this study, we have extended the capability of fluorescence spectroscopy to a spatially-resolved and sensitive tool for rapid visualization of degradation depth-profiles of UV-aged PV laminates, thanks to its characteristics of high spatial resolution (hundreds of nanometers) and rapid acquisition (a few seconds to a few minutes for each image). Confocal fluorescence imaging was used to investigate the effects of UV intensity and wavelength on degradation depth-profiles of both EVA encapsulant and PPE backsheet in glass/EVA/PPE laminates aged on the NIST SPHERE (Simulated Photodegradation via High Energy Radiant Exposure) 26 at 85°C/0 % RH with UV exposure.
To validate and better understand the fluorescence results, the depth profiles of optical (yellowness index), chemical (oxidation, UV absorber concentration), mechanical (modulus), and thermal (melting enthalpy) properties of the EVA degradation were also characterized by UV-Vis-NIR microspectrophotometry (MSP), micro-Fourier-transform infrared spectroscopy (Micro-FTIR), quantitative nanomechanical mapping atomic force microscopy (QNM-AFM) and differential scanning calorimetry (DSC), which tend to be more destructive or more time-consuming in sampling. The correlation of fluorescence mapping to these techniques demonstrates its utility as an effective tool for depth-dependent degradation analysis of polymer laminates. Furthermore, the use of varying UV light conditions (intensity and wavelength) provides critical information on the light-induced nonuniformity and great insight into the mechanisms of degradation with the presence of UV light, which laid crucial groundwork for establishing the correlation between degradation under different laboratory UV-light sources and solar exposure. 27,28 2 | EXPERIMENTAL*

| Materials
Glass/EVA/backsheet laminates (as shown in Figure 1 (a)) made of commercial materials with dimensions of (180 mm × 180 mm) were designed for this study. The glass, with 3 mm nominal thickness, was polished fused silica with a transmittance of around 95% in the UV and visible range. The encapsulant was EVA Photocap 15420P/UF (STR), which contained a UV absorber and curing agent and a vinyl acetate content of 33 % mass fraction. The backsheet consisted of PPE, layers of PET (poly (ethylene terephthalate)), PET and EVA, respectively. The PPE backsheet was composed of pigmented PET outer layer, PET core layer and multiple EVA layers (Coveme). The laminates were prepared in a vacuum environment using a commercial process with a plate temperature of 150°C, evacuation time of 4 min, then pressing time of 1 min and crosslinking time of 13 min. Based on the topography image of the cross-section of the EVA/PPE backsheet laminate (shown in Figure 1 (b)) collected by laser scanning confocal microscope (LSCM) described later, the thickness of the EVA encapsulant layer was determined to be (920 ± 6) μm, and the entire PPE backsheet laminate to be (300 ± 4) μm. The backsheet (labeled PPE) was composed of five layers, from the EVA encapsulant side: an EVA outer layer (25 μm thick), pigmented EVA middle layer (55 μm), EVA inner layer (25 μm), PET core layer (126 μm), and PET *Certain commercial products or equipment are specified in this paper to adequately describe the experimental procedure. In no case does such identification imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that it is necessarily the best available for the purpose outer layer (55 μm). In addition, there were two polyurethane-based adhesive layers between the EVA inner and PET core layers, and PET core and PET outer layers, with thicknesses of 8 μm and 6 μm, respectively.

| Accelerated aging
Accelerated aging was performed with the glass-side of the laminates facing the UV light source in the NIST SPHERE at 85°C/0 % RH under different light intensities and wavelengths for 3840 h. Relatively high temperature was used to simulate harsh climatic conditions of PV modules under hot conditions. 29 For example, the maximum module temperature can reach up to 80°C and 97°C outdoors for rack-and roof-mounted photovoltaic systems, respectively. 30 The NIST SPHERE can produce highly uniform UV irradiance and well-controlled temperature and humidity with a precision of 0.1°C and 0.2 % RH, respectively. 26  were controlled using neutral density (ND) filters with nominal transmittances of 40 % and 60 %, while wavelengths were controlled using bandpass filters with spectral UV ranges of (306 nm ± 3 nm; total bandpass range of 303 nm to 309 nm), (326 nm ± 6 nm; total range 320 nm to 332 nm), (354 nm ± 21 nm; total range 333 nm to 375 nm) and (452 nm ± 79 nm; total range 373 nm to 531 nm). In both cases, filters were placed in front of samples during exposure. The stability of the filters were evaluated during the UV exposure, and essentially no changes were observed in their transmittance spectra (as confirmed in Figures S1 (a) and S1 (b) in the Supporting Information). The irradiance spectrum under each type of filter was also plotted based on the irradiance spectrum from the UV light source and the transmittance from each filter (Figures S1 (a1) and S1 (b1)). The calculated UV irradiances for different ND filters were 60.5 W/m 2 (

| Sample preparation and characterization
Specimen cutting (1 mm x 5 mm) Strips of glass/EVA/PPE backsheet laminate from each exposure condition were cut from the larger laminate using a diamond saw. These were taken from the center of the laminates, to minimize oxygen diffusion effects from the sides. Then the EVA/PPE layers were mechanically peeled off from the glass with tweezers (as shown in Figure 1  For characterization by GC/MS, slices of the EVA encapsulant were microtomed parallel to the exposed surface, as shown in Figure 2

Micro-Fourier-transform infrared spectroscopy (Micro-FTIR) Infrared
analyses were performed by a Thermo Scientific Nicolet iN10 MX infrared imaging microscope purged by dry air, equipped with a liquid nitrogen-cooled mercury-cadmium-telluride (MCT) detector. Chemical changes in the glass-side surface of the EVA encapsulant were measured in Attenuated Total Reflection (ATR) mode using a germanium conical tip with an aperture size of 40 μm x 40 μm. ATR-FTIR is used as its penetration depth into the sample is typically between 0.5 and 2 μm, thus it is sensitive to surface changes compared with transmission mode FTIR. The depth profile of chemical changes of the EVA encapsulant across the cross-sectional direction was also measured in transmission mode using aperture size of (10 μm x 10 μm) with step size of 10 μm on the 5 μm thick cross-sectional slice (as shown in Figure S2 (a)). Spectra were recorded in the spectral range from 4000 cm -1 to 650 cm -1 at 8 cm -1 resolution with 64 scans. Spectra were then processed with a baseline correction and normalization to the peak at 2851 cm -1 assigned to the methylene (-CH 2 -) symmetrical stretching . 33 To ensure reproducibility, at least three different locations were measured for each data point.

UV-Vis-NIR microspectrophotometry (MSP)
UV-Vis spectra were recorded in transmission mode using a microspectrophotometer (CRAIC Technologies, Model MSP 121) in the UV-Vis range between 250 nm and 800 nm across the sample. The microscope objective was a Davin reflecting objective 36x, NA 0.5 with aperture size of (7.9 μm × 7.9 μm). UV-Vis spectra were collected along the crosssectional direction from the exposed side to the back-side with a step size of 7.9 μm, as shown in Figure S2 (a). A cross-sectional slice of only EVA encapsulant, with a thickness of 2 μm, was mounted between quartz microscope slides and cover slips with a glycerin mounting medium. The corresponding yellowness index was calculated based on these spectra according to ASTM E308. To ensure reproducibility, at least three different locations were measured for each data point.   (100 %). The gradual drop in the fluorescence intensity from the exposed surface across the thickness is most likely due to the attenuation of the UV light through the EVA encapsulant. Among the three UV light intensities, the specimen exposed under 60 % UV light has the highest fluorescence intensity near the surface, and reasons for this behavior will be discussed in the next paragraph. Substantial changes in the fluorescence intensity occur within 200 μm from the exposed side of the encapsulant, exceeding which level the intensity reaches the plateau gradually. The overall degradation depth under different UV light intensities is essentially similar, as confirmed by the normalized line profiles given in Figure S3. Figure 4 shows the depth-dependent fluorescence spectra of the same regions shown in Figure 3. A broad fluorescence peak is observed at the exposed side of the EVA, which progressively shifts toward shorter wavelengths and decreases in intensity. No discernable peak is found in the fluorescence spectra of the fresh sample with very low fluorescence intensity (as shown in Figure S4). For 40 % UV light, the peak position shifts from 600 nm to 520 nm about halfway into the EVA layer, and no obvious shifts are observed thereafter.

| Fluorescence imaging for samples aged under different UV light intensities
For 60 % UV light, the emission peak position shifts consistently from 617 nm to 520 nm throughout the entire EVA layer. And for 100 % UV light the peak position (around 635 nm) changes very slowly at first, and then a sharp shift to 520 nm is observed at about 100 μm.
Also note that the peak position of the exposed side of samples under  that the fluorescence intensity of the 60 % UV sample is higher than the 100 % UV sample. The reason for this drop for 100 % UV light could be due to secondary reactions related to the short chain α, β-unsaturated carbonyl species with more severe UV exposure. A similar peak intensity drop for severely discolored EVA encapsulant was also reported by Pern. 20  is only about 9 % below 340 nm while increases sharply above about 380 nm (as shown in Figure S5). Although the light centered at 452 nm is also strongly reflected, only a small increase in the fluorescence intensity on the backsheet side of the EVA encapsulant is observed.

| Fluorescence imaging for samples aged under different UV light wavelengths
It is known that the degradation efficiency of light at this wavelength range is rather low, and more importantly, light with longer wavelengths could bleach the chromophores and reduce yellowing and fluorescence. [37][38][39] The competitive reactions between formation and depletion of chromophores during degradation may be the reason that no obvious increase in the fluorescence intensity at the backsheet interface was observed for samples aged under broad band UV spectrum with neutral density filters (Figure 3).
The effect of UV wavelength on the rate of decay in fluorescence intensity as a function of depth can be clearly seen from the normalized curves ( Figure 5 (f)), that is, a higher attenuation rate at a shorter wavelength. This relationship can be quantified by Lambert-Beer law, 40,41 in which light intensity I(λ,x) decays exponentially in the material with the penetration depth x: where I 0 is the incident intensity, α(λ) is the absorption coefficient. The drop in fluorescence intensity F (λ,x) is then fitted using an exponential relationship No obvious change in the shape of fluorescence spectra as a function of depth is observed (as shown in Figure S6). Compared with the degradation of samples under different light intensities, the degradation process under different wavelengths is still in the early stage, so the shift in peak position of the fluorescence spectra has not occurred yet.

| Correlation between fluorescence and other material properties
The above results have demonstrated that fluorescence imaging is an effective and spatially sensitive tool to visualize the depth-dependent non-uniformity in degradation of the aged encapsulant EVA in the UV-  (shown together later in Figure 9), indicating that their mechanisms of degradation could be related.
An important origin for the discoloration and fluorescence of EVA encapsulant is the interactions among stabilizing additives, peroxides and polymers. [44][45][46][47] Specifically, the interaction between the peroxides, phosphite thermal stabilizer or radicals from degradation of polymers with the benzophenone based UV absorbers could lead to chromophore formation. 45 A recent interlaboratory study by Miller et al. 44 shows that the UV absorber is mostly responsible for the discoloration of EVA, as no obvious discoloration is found for EVA encapsulant without UV absorbers. As shown in Figure 6  ATR-FTIR spectra of fresh and exposed EVA on the surface (glass-side) are shown in Figure 7 (a). The main peak of interest for degradation of EVA is at 1740 cm -1 , which is assigned to ester or aldehyde groups. 48 There is no obvious change in intensity for this band, which could be due to the competition between destruction of esters and formation of aldehydes. Nonetheless, a broadening of this band is observed and absorption shoulders around 1715 cm -1 and 1780 cm -1 appear, indicating the formation of other carbonyl groups such as ketones and lactones, respectively. Ketones are formed either by Norrish III reaction or by decomposition of hydroperoxides, while formation of lactones is related to the back-biting process in the vinyl acetate groups. 33,49 There is a strong depth dependence for carbonyl groups formation (as represented by the absorbance at 1715 cm -1 ) in the aged EVA encapsulant, shown in Figure 7 (b). The carbonyl products decrease rapidly further from exposed surface, and the degradation depth is consistent with the evolution in the fluorescence and discoloration, showing that the kinetics of EVA degradation and chromophore formation are similar. Carbonyl products from the degradation of EVA can be conjugated with double bonds, and the resultant α,β unsaturated ketone contributes to the discoloration and fluorescence, which has been discussed widely in the literature. 23,50,51 By comparing with model compounds, Rodríguez-Vázquez et al. 51 proposed that the discoloration of EVA is not due to polyene formation, while it is linked to α,β-unsaturated carbonyl species. The relative concentration of carbonyl products reaches a stable level at a depth of about 200 μm.
The Derjaguin-Muller-Toporov (DMT) modulus of the EVA encapsulant before and after aging across the thickness is measured by QNM-AFM (Figure 8 (a)). The modulus of the EVA encapsulant is an important degradation indicator, 52  it is an ideal alternative to quantify degradation related changes.

| Depth-dependent degradation of PPE backsheet layers in the glass/EVA/backsheet laminate
The capability and sensitivity of fluorescence imaging in depthdependent degradation characterization of PV laminates can be further demonstrated by the study of the PPE backsheet in the UV exposed glass/EVA/PPE laminate. Figure 10 shows  induced. This hypothesis is further supported by the similar fluorescence intensity observed from the adhesive layers in backsheet aged under different wavelengths (as shown in Figure S7). There are no obvious changes in the PET core layer. These results provide the direct evidence on the degradation of the EVA middle layer and two adhesive layers, indicating that the instability of those layers, and potential inner layer 58 and interfacial failures in the backsheets during laminate exposure.
The degradation of the adhesive layers has been further confirmed by AFM-QNM of the PPE film in the glass/EVA/PPE samples exposed at UV/85 ○ C/0 % RH under 100 % UV ( Figure 11). For the fresh sample, small nodule structures in each adhesive layer are found, which are dispersed uniformly in the matrix and the adhesion of nodule structures is relatively lower than the matrix material. The interfaces between the adhesive and the adjacent layers are sharp and clear.
After exposure to 100 % UV, for the EVA/PET adhesive layer ( Figure 11 (a)), the size of the small nodules basically remains unchanged, while the interfaces with the EVA and PET layers is less defined. For the PET/PET adhesive layer (Figure 11 (b)), the small nodular structures observed in the fresh sample become larger, probably due to the coalescence of the small latex particles under heat, and a groove-like structure appears after aging, indicating the loss of some material. These results indicate the degradation of both adhesive layers has taken place, leading to microstructural and mechanical changes during exposure, which was previously evidenced by their high fluorescence emission. These results also suggest a potential risk of inner layer failures within the backsheet, which could raise safety concerns for electrical insulation, and ultimate reliability issues of the PV modules.

| CONCLUSIONS
This study has demonstrated that fluorescence imaging is an effective and spatially sensitive technique to monitor depth dependent degradation, which is important for failure mechanism analysis for PV mod- 3. The modulus and melting enthalpy do not closely match the fluorescence profiles, but follow a similar trend between themselves, suggesting a relationship between the thermal/structural and mechanical properties across the depth of the EVA encapsulant; 4. Fluorescence imaging has also showed the obvious degradation of the EVA middle layer and two adhesive layers in the PPE backsheet, suggesting a potential risk of inner layer failures within the backsheet, which could raise potential safety concerns and ultimate reliability issues of the PV modules.
In summary, fluorescence imaging has enhanced our ability to visualize heterogeneous degradation in-depth and opened a new route for the understanding of the gradient degradation of each layer in multilayer PV packaging materials, which would unlock new opportunities in rapid failure assessment for whole PV systems.