UV-induced degradation of high-efficiency silicon PV modules with different cell architectures

Degradation from ultraviolet (UV) radiation has become prevalent in the front of solar cells due to the introduction of UV-transmitting encapsulants in photovoltaic (PV) module construction. Here, we examine UV-induced degradation (UVID) in various commercial, unencapsulated crystalline silicon cell technologies, including bifacial silicon heterojunction (HJ), interdigitated back contact (IBC), passivated emitter and rear contact (PERC), and passivated emitter rear totally diffused (PERT) solar cells. We performed UV exposure tests using UVA-340 fluorescent lamps at 1.24 W (cid:1) m (cid:3) 2 (at 340 nm) and 45 (cid:4) C through 4.02 MJ (cid:1) m (cid:3) 2 (2000 h). Our results showed that modern cell architectures are more vulnerable to UVID, leading to a significant power decrease ( (cid:3) 3.6% on average; (cid:3) 11.8% maximum) compared with the conventional aluminum back surface field (Al-BSF) cells (< (cid:3) 1% on average). The power degradation is largely caused by the decrease in short-circuit current and open-circuit voltage. A greater power decrease is observed in bifacial cells with rear-side exposure compared with those with front-side exposure, indicating that the rear side is more susceptible to UV damage. Secondary ion mass spectroscopy (SIMS) confirmed an increase in hydrogen concentration near the Si/passivation interface in HJ and IBC cells after UV exposure; the excess of hydrogen could result in hydrogen-induced degradation and subsequently cause higher recombination losses. Additionally, surface oxidation and hot-carrier damage were identified in PERT cells. Using a spectral-based analysis, we obtained an acceleration factor of 5 (cid:5) between unpackaged cells (containing a silicon nitride antireflective coating on the front) in the UV test and an encapsulated module (with the front glass and encapsulant blocking 90% of the UV at 294 nm and 353 nm, respectively) in outdoor conditions. From the analytical calculations, we show that a UV-blocking encapsulant can reduce UV transmission in the module by an additional factor of (cid:6) 50.

aluminum back surface field (Al-BSF) cells (<À1% on average).The power degradation is largely caused by the decrease in short-circuit current and open-circuit voltage.A greater power decrease is observed in bifacial cells with rear-side exposure compared with those with front-side exposure, indicating that the rear side is more susceptible to UV damage.Secondary ion mass spectroscopy (SIMS) confirmed an increase in hydrogen concentration near the Si/passivation interface in HJ and IBC cells after UV exposure; the excess of hydrogen could result in hydrogen-induced degradation and subsequently cause higher recombination losses.Additionally, surface oxidation and hot-carrier damage were identified in PERT cells.Using a spectralbased analysis, we obtained an acceleration factor of 5Â between unpackaged cells (containing a silicon nitride antireflective coating on the front) in the UV test and an encapsulated module (with the front glass and encapsulant blocking 90% of the UV at 294 nm and 353 nm, respectively) in outdoor conditions.From the analytical calculations, we show that a UV-blocking encapsulant can reduce UV transmission in the module by an additional factor of $50.

| INTRODUCTION
After years of improvement in photovoltaic (PV) module performance, including the reduction of power degradation rates toward a mean of À0.5%Áyear À1 to À0.6%Áyear À1 for crystalline silicon (c-Si) technology, 1 there are new pieces of evidence that the degradation rates for many c-Si modules are now increasing.For example, Trina Solar published an approximate À4.5%Áyear À1 degradation rate for multiple module types mounted in Singapore, and the Department of Energy (DOE) National Laboratory Regional Test Center program showed degradation of À1%Áyear À1 to À2%Áyear À1 for various types of high-efficiency c-Si modules, even after initial light-induced degradation (LID) stabilization. 2 The increased initial PV performance of next-generation high-efficiency solar cells has been attributed to the high-quality wafers and their longer carrier lifetimes.Improvements in carrier lifetimes are due in part to improved surface passivation.However, the recent increase in power degradation rates of modern cell types is concerning and needs to be better understood and addressed if c-Si modules are to reach a 50-y module lifetime.
Historically, the harmful effects of UV radiation have largely been associated with the aging of module packaging materials and have led to encapsulant discoloration, 3,4 delamination, 5,6 and backsheet cracking. 7,80][11][12] To avoid carrier recombination at defects and improve minority carrier lifetimes, the cell surfaces are generally passivated using dielectric layers, such as hydrogenated silicon nitride (Si x N y :H) and/or silicon dioxide (SiO 2 ).However, UV exposure can disrupt this surface passivation by damaging the passivation layer itself or the passivation layer/Si cell interface [11][12][13][14] as well as causing subsurface damage in the silicon. 15,16[19][20][21][22][23][24] Emerging higher performance cell technologies, such as p-type passivated emitter and rear contact (p-PERC) and n-type passivated emitter rear totally diffused (n-PERT) cells, have been observed to be more sensitive to UV radiation. 21Additionally, as bifacial modules gain market share, it is important to consider that both the front and back of the bifacial cell are potentially susceptible to UV-induced degradation (UVID). 20UVID can be characterized by a gradual loss in module performance.For example, JinkoSolar published an efficiency loss of À4% to À7% in industrial solar cells after exposure to approximately 540 MJÁm À2 (150 kWhÁm À2 , i.e., 25 sunny days AM1.5G) of UVA light. 25In some PERT cells, a UV dose of 1.8 GJÁm À2 at 40 C resulted in 15% power loss. 21We note that the UVID described here results from the surface region and/or interface effect(s)-distinct from the typical LID (a fast initial power decrease attributed to the evolution of boron-oxygen and boron-iron complexes in the bulk of silicon) 26,27 and light and elevated temperature induced degradation ("LETID," a slow progressing degradation activated during light exposure or via electrical current at temperatures above 60 C). 28,292][13][14] reported an increase in recombination centers at the Si/SiO 2 interface due to UV light photoinjecting electrons from the silicon conduction band into silicon dioxide.UV exposure is effective in disrupting surface passivation by changing the fixed charge density and increasing the interface defect density near the Si surface.30,31 One explanation for the vulnerability of Si x N y passivation arises from the deposition processes of Si x N y films.Transmission electron microscopy (TEM) examination revealed that during plasma-enhanced chemical vapor deposition (PECVD), a damaged layer of around 50 nm is formed at the c-Si side of Si x N y /Si interface, which is passivated with H atoms. 15 However, the recombination velocity at this region is increased by the UV irradiation, resulting in a decrease in carrier lifetime.Another paper revealed that subsequent etching and removal of the damaged passivation and defect layer tend to recover the carrier lifetime.16 Further, the defects in the damaged layer were electrically inactivated due to high-temperature annealing (600 C) and thereby resulting increase in the effective lifetime.32 The generation of hot carriers (either electrons or holes) is another proposed UV damage mechanism.12 For example, hot electrons can be generated when a high-energy photon (i.e., UV radiation) excites an electron out of the valence band and beyond the conduction band.Hot carriers are extremely mobile and have sufficient energy to surpass an interfacial potential barrier, allowing them to be injected across the interface when the carrier concentration exceeds 1.5 Â 10 17 cm À3 .10 The hot carriers generated in Si cells can have sufficient kinetic energy to damage the passivation layer and increase the interface state density.Hot carriers are more prevalent in PV devices that operate under the extraterrestrial solar spectrum, concentrated sunlight, or high electrical current.There are many reports in the literature concerning the effects of oxidation and other processing variables on hot-electron damage. 33 Fo instance, texturization, trichloroethane oxidation, and an aluminum anneal all contribute to the instability of the passivation interfaces.Dry thermal oxides (hot-carrier resistant) or poly-Si (block water) can provide good protection from hot carrier damage.34 Once damaged, it may be possible to anneal at 250-300 C to restore the correct charge state.There is evidence that at least some of the generated interface defects are hydrogen-related.Gruenbaum et al. 9 found that the defects could be recovered fully by annealing in a hydrogen-containing atmosphere but only partially in nitrogen.Buchanan et al. 35 correlated the electron injection and generation of interface traps with the release of trapped hydrogen in the oxide and its migration to the interface.
They inferred a threshold electron energy of $2 eV in excess of the oxide conduction band for both the hydrogen release and defect generation.
Previous studies suggest several possible degradation mechanisms associated with UV exposure, but the UV stability of modern cell types is not well understood.The current manuscript focuses on identifying the UV susceptibility of a variety of contemporary PV cells, distinguishing between their photovoltaic performance, composition, and optical performance through UV aging.We sought to identify the most affected types of cells and differentiate the degradation mechanisms enabling UVID.Understanding the damaging effects of UV radiation in emerging silicon solar cell technologies will enable the identification of the underlying mechanisms that may affect both the power output and durability of modules.

| Test samples
We carried out the experiment on a variety of unencapsulated, commercial c-Si solar cell architectures, as outlined in Table 1.The test samples are grouped by cell technology; different cell makes are designated with a letter index (A-L).The sample set includes highefficiency silicon cell technologies, such as heterojunction (HJ), interdigitated back contact (IBC), PERT, and PERC cells.We compared the performance of these cells against the conventional legacy aluminum back surface field (Al-BSF) cells.The samples used in this study were cut from high-efficiency cells (in the order of 20%, or around 15% for the legacy Al-BSF).Detailed descriptions of cell dicing and tabbing interconnection are described in Section S1 of the Supporting Information.Both mono-and multi-crystalline cells were examined.The set also includes bifacial cells, which were irradiated from both the front and rear sides to study the UV sensitivity of each face.In the case of bifacial IBC cells, the rear side features an open metallization grid where UV light is transmitted/absorbed in the narrow gap between p and n connected fingers.At least three replicas of each cell type were chosen to increase the confidence in the average results and to identify outlying measurements.
The test samples were mounted on in-house designed fixtures at NREL, made of polyoxymethylene for protection and handling of the cells throughout UV testing.The test fixture featured a black mask to avoid optical reflection to the cell from the surrounding white surfaces, as shown in Figure 1A.SLAC used the chamber manufacturer's standard specimen holders, which are made of clear anodized aluminum, as displayed in Figure 1B.The base surface of the holder was masked with polyimide tape to provide electrical insulation to the cells.   Assuming an ideal sunny day at solar noon, as in Habte et al., 37 we converted the chamber UV dose at 340 nm (equivalent to 4.02 MJÁm À2 ) to the measured outdoor cumulative exposure from the measured wavelength range of 295-385 nm (averaged from years 2000-2020) using the Simple Model of the Atmospheric Radiative Transfer of Sunshine (SMARTS) code. 38,39Because the controlled chamber dose was converted relative to the measured outdoor dose, the daylight hours and weather history (cloudiness) were also considered.

| Characterization
To investigate degradation in cells after UV exposure, we employed several characterization methods, including current-voltage (I-V) testing, spectral reflectance measurements, secondary ion mass spectroscopy (SIMS), and X-ray photoelectron spectroscopy (XPS).I-V and spectral reflectance measurements were used to examine the changes in the electrical and optical performance of the cell, respectively.SIMS was primarily used to measure the hydrogen distribution at the interface and bulk of Si, whereas XPS was used to examine the surface chemistry.Detailed descriptions of these characterization methods are provided in Section S2.

| Spectral optical modeling of UVID
To better quantify the UV dose for the unpackaged cells used in this study relative to a PV module, as well as to help diagnose the degradation mechanism for UVID, we developed an optical model for c-Si modules containing Si x N y antireflective coating on the front side (excluding HJ cells).While the analysis and verification of the optical effect of the indium tin oxide (ITO) layer of HJ cells are outside the scope of this work, the measured reflectance of the cell surfaces may be used as a first estimate to compare the UV irradiance at the Si x N y /Si interfaces on the cell surfaces (Figure S1).
Because the optical performance of antireflective coatings may vary significantly with UV wavelength and also between makes of cells, future study is recommended.The spectral absorptance of each module layer and the spectral reflectance of each interface were evaluated so that a transfer function could be determined for the incident optical flux at the Si x N y layer and c-Si cell surfaces.All spectra were evaluated in 1-nm increments, from at least 250 nm to 2500 nm.Thin-film layers that were considered include the antireflective coating on the front glass, AR g , and the antireflective coating on the cell surface, AR c .The optical performance was evaluated using the approach in Miller. 40Briefly, the transmittance (τ i , dimensionless) and reflectance (ρ i , dimensionless) of individual components (e.g., glass) were measured so that the complex refractive index could be iteratively evaluated to a converged solution.The optical performance of coupons (e.g., AR g /glass, glass/encapsulant/glass, and glass/encapsulant/cell) was measured to assess material stacks or bulk materials (encapsulant).Bulk/thin-film/bulk material stacks (e.g., air/AR g /glass and encapsulant/AR c /Si) were compared with an analytic model. 41The complex refractive index of a representative Si x N y AR c thin film was determined using ellipsometry.All samples were examined in an unaged condition.The complex refractive index of Si 42,43 and the terrestrial, real refractive index of air 42 were taken from the literature.The source spectra for the NREL custom chamber used in this study were verified using an OL 756 spectrometer with an OL IS-670 integrating sphere (Optronic Laboratories, LLC).Once the refractive index was known, the absorptance was determined from the Beer-Lambert law, and the reflectance was determined from Snell's law.The transfer function (F t ) for the cell or module was then determined from the product of the reflectance (ρ) and absorptance (α) events occurring along the optical path (see Equation 1).
In this section, we first compare the UV susceptibility of different modern unencapsulated cell types in a UV screening experiment.
Section 3.2 presents the results from multiple characterization techniques, including I-V, SIMS, XPS, and optical reflectance.The extensive chemical and optical characterizations were employed on select cell makes to diagnose the degradation mechanism(s) for the cells showing the greatest power degradation from UV exposure.
Because only a subset of the samples was examined in detail, the diagnosis may not apply universally within a given cell technology, and the variation between makes of cells cannot be assessed.

Section 3.3 describes the test results of cells under open-circuit and
short-circuit conditions and the associated hot-electron damage in the cells.

| UVID in unencapsulated cells
Figure 2 shows the percent change in maximum power (P max ) of unencapsulated cells with front surfaces irradiated over a 2000-h UV exposure test at NREL and SLAC.The statistical confidence in measured P max is shown here, while that of other I-V parameters is shown in Figure S2.The greatest P max degradation was observed in the HJ cells, followed by the n-PERT and p-PERC cells, whereas the power decrease was least pronounced in their Al-BSF counterparts.
The time evolution of P max degradation in most of the cell types follows a linear trend, except for "A HJ " and "D PERT ," which show a nonlinear or exponential decrease.We determined the P max degradation rate by assuming an initial linear power reduction, as in Figure 2 (also reported in several studies 44,45 ).We find that HJ cells suffer from the greatest power loss, of À5.46%Áyear À1 (average), with a maximum degradation rate of À8%Áyear À1 .The P max of p-PERC and n-PERT cells degrades at an average rate of À1%Áyear À1 , whereas the P max of Al-BSF cells degrades at À0.36%Áyear À1 .IBC cells showed a different behavior: the cells exhibited a slight increase in power.Note that the rapid power loss observed in our UV stress testing was obtained on the bare (unpackaged) cells.The equivalent power loss in a packaged module was determined using a spectral analysis, which is presented in Section 3.4.Further, the degradation trajectory varies for different cell types within the same PV technology, indicating that the degradation mechanism is influenced by more than just cell technology, and depends on the materials and manufacturing processes used.
Figure 3 shows the final degradation after 2000 h of UV exposure in every cell type, including the bifacial cells (indexed as +).
Most of the bifacial cells with rear-side exposure exhibited a greater P max decrease than those with front-side exposure, indicating that the rear side of the bifacial cell is more vulnerable to UV radiation.
This could result from differences in the surface field and passivation schemes used on the front versus rear sides, making the rear side more susceptible to UV degradation.The average P max degradation rate observed on the rear side of bifacial cells is À2.4%Áyear À1 , in contrast to À1.4%Áyear À1 on the front side.Note that the total UV dose on each side of the bifacial cells under testing was the same; however, during actual field operation, the back surfaces of bifacial modules receive only a fraction of UV radiation compared to the front surfaces, depending on the mounting position, ground albedo, and presence of UV absorbers in the module packaging materials.Although not specifically examined here, the number and thickness of the layers may vary between the front and back of the cells, readily affecting the UV dose within the AR c and its interface with the cell.
To see the variability of UVID on each cell type and its correlation with the governing characteristic power loss, we studied the cell electrical parameters separately.Figure 3

| HJ cells
Figure 3 shows that the average rapid power loss observed in HJ cells after UV exposure on the front side is À11%.This power loss is dominated by a decline in V oc (À3%) and FF (À7%) while the I sc decrease (À1%) contributes the least.Similarly, the average degradation observed on the rear side of HJ cells in P max , V oc , FF, and I sc is À14%, À6%, À3.5%, and À1%, respectively.A significant increase in series resistance (R s ) is observed from the I-V curve slope at V oc (shown in Figure 4A), which could be attributed to an increase in sheet resistance of the ITO front electrode or deterioration of the ITO/a-Si and (p)a-Si/(i)a-Si passivating interfaces.No obvious signs of delamination, corrosion, intermetallic reaction, or grain growth were observed for the gridlines in optical microscopy (Figure S3).The comparison of dark I-V curves before and after UV exposure indicated an increase in R s as well as in the diode ideality factor and recombination current (Figure S4).Similarly, Jordan et al. reported the non-linear power loss in 10-year-old fielded HJ modules, where most degradation occurred in the first 2 years and was attributed to significant V oc losses and a small decrease in FF.The V oc decrease was ascribed to an increase in carrier recombination, whereas FF degradation is primarily caused by an increase in R s. 46 It is important to note that the HJ cell examined was the most affected of the specific manufacturer in this study; therefore, the observed power degradation may not apply to all cells independent of cell structure.
UVID was previously associated with hydrogen redistribution across the passivation layer and the silicon bulk in the cell. 47To quantify H distribution as a result of UV exposure, we used SIMS, as this technique offers good sensitivity in detecting H and other trace elements.Figure 5A  Finally, to further explore the source of UVID in the HJ cells, we performed optical reflectance measurements on the front surface of cells.Figure 5D shows that the difference in spectral reflectance between the unaged and UV-aged A HJ cell is negligible (Δρ rsw of 0.3%), suggesting that the front coatings are not significantly deteriorated.Combining the characterization results, we postulate that the damage to these HJ cells originates from the hydrogen effusion near the surface with subsequent migration into the bulk silicon, resulting in increased recombination loss.

| IBC cells
IBC cells display a small increase in power under both front-and rearside exposures (2% and 6%, respectively, with uncertainty in the results of ±2% and ±4%, respectively) after UV aging, despite losses in I sc and V oc (À1% to À2%).This anomalous behavior is attributed to the increase in FF (+3% and +8%, respectively).The reason for the increased FF is unclear, but it could be an artifact of flash I-V testing or a result of changes in charge state induced by UV light.We did not explore the possible metastability of this performance gain in this work.
The SIMS profiles of the B IBC cells are shown in Figure 7B.The H profiles (red) show an increased H concentration at or near the Si x N y /SiO 2 interface and underneath the Si bulk after UV aging.It has been hypothesized that the energetic UV photons break the weak Si-H bonds, releasing hydrogen, which then diffuses toward the inner layers. 47Hydrogen is crucial for defect passivation, but excess H atoms could result in hydrogen-induced degradation. 47By this mechanism, I sc and V oc losses can be attributed to extra H atoms forming Hclusters that cause increased carrier recombination.No significant differences were seen in the distribution of Si, O, and N profiles for the IBC cell.Similarly, the XPS spectra from the Si 2p, N 1s, and O 1s profiles (shown in Figure S6) demonstrated no significant differences after the UV exposure.Further, no major change was observed in the reflectance spectra at any wavelength.Taken together with the characterization results, we infer that the cause of damage to these IBC cells originates from the excess of H in c-Si.For these IBC cells, which are made of high-quality wafers, extra H atoms cannot further passivate defects, but do cause more degradation and compromise the performance.

| PERC and PERT cells
An interesting observation arises in p-PERC cell technologies.There is considerable variation in the extent of UVID in bifacial p-PERC cells between the front and rear sides.In these cells, the P max decrease under front-side exposure lies in the narrow range of À1% to À4%, but reaches up to À13% under rear-side exposure.The factors affecting the P max decrease in p-PERC cells with rear-side exposure are I sc and V oc losses.The highest I sc degradation (À10%) and V oc loss (À3%) are seen in J PERC cells.This is because bifacial p-PERC cells typically adopt different passivation schemes on the front and rear sides.The front surface of a bifacial PERC cell includes a selective phosphorous emitter (greater doping beneath the metal contacts) and a thin thermal oxide SiO 2 passivation (ensuring low carrier recombination velocity), whereas the rear surface includes dielectric passivation of the AlO x / Si x N y stack.We hypothesize that the lack of back surface field on the rear surface makes it more vulnerable to UV, as the rear instead relies on dielectric passivation from built-in charges within the AlO x layer.
Another important factor is the difference in optical performance (reflectance and absorptance, resulting from a different thickness and refractive index) of the hydrogenated Si x N y layer, which gives a different optical transfer function for the back of the cell.Generally, the Si x N y thickness on the rear side is greater than on the front side to improve passivation, 51 which would increase the absorption of UV photons.The variation in the refractive index also plays a vital role in cell degradation under UV exposure.Witteck et al. reported an increase in power loss with decreasing refractive index of the Si x N y layer during UV exposure. 18It is evident that the Si x N y layer absorbs significant amounts of short-wavelength light, especially for films with a higher refractive index. 52Typically, the rear Si x N y layer of the cell has a lower refractive index than the front Si x N y layer, thereby reducing reflectance and increasing the UV flux to the rear side.The concentration of H in the Si x N y layer, which may contribute to degradation, also varies with the refractive index. 53The net effect of thickness and refractive index remains to be evaluated.
As shown in Figure 3, n-PERT cells from different manufacturers exhibited power degradation to varying extents after UV exposure.
The P max drop was greatest in D PERT (À7%) and C PERT (À3%), whereas the E PERT and F PERT cells showed a slight improvement in power (2%).
All of these cell types contain similar front passivation structures, thus, the variation in their UV sensitivity could be attributed to factors like differences in passivation layer deposition steps or the dopant concentration profiles of the front emitters.Furthermore, the bifacial D PERT cells exhibited greater power degradation when irradiated on the rear side.The greater power loss is mostly correlated with the subsequent decreases in I sc and V oc , whereas the FF decrease is negligible, as shown in  Figure 5D shows a nearly 1% increase in the representative solarweighted reflectance of photon irradiance (ρ rsw ) of the aged D PERT sample, occurring mostly in the spectral range of 240-500 nm.The ρ rsw is similar to the representative solar-weighted characteristic for transmittance in Miller et al. 48and can be expected to give a $1% decrease in the photocurrent generation.This is approximately half of the I sc loss (À2.3%) and the subsequent P max loss in the D PERT cell (Figure 3).Based on the XPS data, the change in ρ rsw is caused by chemical changes near the Si x N y passivating layer/Si interfacial region.
By comparing composition and optical performance, we observed both the chemical and physical changes causing the current loss and increased recombination activity at the Si/passivation interface.

| UVID and hot-electron damage
The UV testing was carried out at $2.5Â UV suns, that is, modestly concentrated sunlight that may enable hot-electron damage in the cells, as mentioned in the introduction section.PV installations, however, are operated at 1Â UV sun, which may limit hot-electron damage.To study the influence of electron photoinjection on hot-carrier damage, select cell technologies (HJ, n-PERT, and p-PERC) were subjected to UV irradiation from the front side of the cell in a short-circuit configuration.The selection of the cells for the open-circuit and short-circuit studies was undertaken prior to starting the experiment, a few criteria such as availability of cells and variety of cells in the test, were considered.

| UVID in PV modules using spectral optical analysis
We used a spectrum-based analysis, where the absorption of each layer and the reflectance 59  Next, we estimated the UV dose for the UV screen test on bare cells relative to that of a PV module operating outdoors.Table 2 provides the UV intensity at the wavelength of 340 nm, incident at either the encapsulant/Si x N y interface or the Si x N y /Si interface.Values are given for a bare cell and a PV module encapsulated with a UVtransparent or a UV-blocking POE. Figure 9 and Table 2 remind us that the UVA-340 lamps produced higher irradiance levels relative to the terrestrial sun below about 370 nm.Comparing the UVA intensity of the lamp and the sun at 340 nm confirms an acceleration factor (AF) of 2.5 on a WÁm À2 basis in our UV screen test relative to AM1.5G.The spectral analysis, which is based on absorbance/ reflectance interactions of incident flux with different material layers of a PV module (Figure 9), shows that the UV intensity at the Si x N y layer is attenuated by a factor of 2 in a module compared with a bare cell.Together, the intensity and transfer of radiation give a net AF of $5Â, which may be used to compare the degradation of the bare cells in our study to the expected results for a fielded module.Using the net AF, the average degradation rate of À0.6%Áyear À1 obtained in the UV aging test on unpackaged cells is translated to À0.12%Áyear À1 for an encapsulated module in outdoor conditions, with a corresponding maximum degradation rate of À0.73%Áyear À1 .This is comparable to aggregated published mean degradation rates of c-Si PV modules, which are in the range of À0.5%Áyear À1 to À1.1%Áyear À1 .The analytical calculations in Table 2 also show that UV-blocking encapsulants can reduce UV transmission by an additional factor of $50.The aforementioned analysis does not consider wavelength sensitivity-for example, if the UVID degradation has a nonlinear action spectrum for damage. 40

| POTENTIAL UVID MITIGATION SOLUTIONS
The negative impact of UVID can be minimized through changes in module packaging materials and/or cell design.Historically, PV encapsulants were formulated with suitable additives of varying UVblocking abilities.Industry product formulations such as benzophenone-or benzotriazole-based UV absorbers have their characteristic absorption peaks at wavelengths of 280-380 nm, 60 which allow photon wavelengths greater than 360 nm to reach the cell surface.Recent advancements have led to an improved understanding of modern encapsulant formulations, which allow for a more tailored UV cut-off wavelength compared to old formulations.For example, triazine-based additives might be used to shift the UV absorptance to the spectral range of 260-280 nm, 32 decreasing the UV cut-off wavelength below 360 nm.Multiple UV absorbers may also be used in the formulation to further tailor the spectral range of optical transmittance of the encapsulant.The raw materials cost of a UV absorber made using industry encapsulant formulations is currently estimated to be in the range of $0.05-$0.2ÁmÀ2 .The consequences of using a UV absorber include possible discoloration of the encapsulant and a reduction in performance from the chromophore species. 61 The acceleration factor is the ratio of UVA intensity at 340 nm at a specific material interface (or stack) in the cell (or module) to that of the sun.
the use of a UV absorber, the cost of the absorber and the resulting loss of electricity production from any potential discoloration it may cause needs to be less than the cost of the loss of electricity production from UVID.
Other solutions for mitigating UVID in encapsulated PV modules include using a specialized glass cover, such as cerium-containing glass, advantageously solarized glass, or antireflective coatings tailored to reduce UV transmittance.Previous studies have confirmed the ability of Ce-doped glass to reduce discoloration of ethyleneco-vinyl acetate (EVA). 62,63The UV-blocking ability of Ce glass is expected to protect solar cells from the damaging effects of UVID.
However, its utility is limited due to energy-intensive processes, the high cost of production, and poor recyclability.The presence of rareearth elemental impurities in the glass makes the entire recycling process less economically viable for the PV industry due to costly reducing agents for dissolution and the complexity of the metal extraction processes. 64A second method to induce desirable changes in the glass material is solarization, in which a glass is exposed to either sunlight or artificial light sources for extended periods.King et al. demonstrated a $2% loss in solar-weighted transmittance (>800 nm) as well as an increase in absorption in the 330-400 nm region of the UV spectrum for 0.2 wt.% Ce-containing float glass after several hours of solarization under highly concentrated, full-spectrum solar radiation above 1000 suns, with sample temperatures maintained below 100 C.
The incident photon dose was equivalent to 8 years outdoors in Golden, Colorado, USA. 65Here, Ce acts as a solarization agent because the Ce 3+ state initially present in the glass is oxidized to the Ce 4+ state under prolonged light exposure.Ce is no longer used in the PV industry, but Mn and Sb are often added to soda-lime glass for whitening and improvement of transmittance. 66The redox reaction between Mn and Sb advantageously results in solarization that reduces UV transmittance, but typically on the order of a few nanometers.Antireflective coatings, which immediately improve electricity generation on the order of a few percent, have found widespread use in PV since around 2005. 67 The currently popular graded-index coatings using porous silica do not affect UV transmittance greatly.However, the destructive interference coatings that use layers of metal oxide films give a greater reduction. 67Ideally, a multilayer coating could be designed to both reduce UV transmittance and reject spurious infrared wavelengths (to improve module performance relative to its temperature coefficient).The adoption of multilayer coatings for PV would depend on cost reduction, widespread use and economies of scale, and/or the need for more durable AR g coatings in soilingprone locations.
On the cell side, modifications to passivation schemes can help mitigate UVID.For instance, the Si x N y antireflective coating with a refractive index of greater than 2.3 was found to decrease the degradation rate. 53With the increase in refractive index, the UV absorption in the Si x N y layer also increases, which in turn reduces the number of UV photons reaching the Si x N y /Si passivation interface.This reduces damage, but also leads to a non-negligible loss in photocurrent, to the extent that the UV absorption continues into the blue wavelengths.
Further, cells featuring AlO x /(p + )Si passivation are more UV-stable than those with a Si x N y passivation layer. 21This could be due to the formation of a stable interfacial SiO x layer between Si and AlO x 31,68 and an increase in negative fixed charges within the AlO x layer under UV illumination. 69These factors improve the passivation, which in turn increases the module power output.The most durable among high-efficiency cell technologies (HJ, IBC, PERC, or PERT) remains to be proven for a wide variety of makes and models of cells.

| CONCLUSIONS
In this study, we demonstrate the UV susceptibility of various modern PV cell designs through an accelerated UV exposure test on unencapsulated silicon solar cells, including bifacial cells.Highefficiency modern cell technologies, including HJ (À11%), n-PERT (À3% to À7%), and p-PERC (À1% to À4%) showed greater P max losses compared to conventional Al-BSF cells (À1%).We found the rear sides of bifacial cells to be more vulnerable to UVID than the front sides, which could be attributed to differences in the surface

F I G U R E 1
Carrier fixtures used at (A) NREL and (B) SLAC chambers, both featuring special masking to reduce the optical reflection and provide electrical insulation to the unencapsulated solar cells Prior to UV exposure tests in the weathering chambers, the test samples were preconditioned in natural sunlight for a cumulative broadband dose of ≥15 kWh to stabilize against boron-oxygen (B-O) LID.The cells were placed under a museum-grade acrylic glass sheet (UF-5, Arkema S.A.; 100% light blocking below 404 nm) to filter the UV radiation.The cell temperature was also periodically monitored using a calibrated thermocouple to ensure it stayed below 45 C, averting the activation of LETID.
shows the percent change in each I-V parameter after 2000 h of UV irradiation.We found that the contributing factors to P max degradation are not the same for all cell technologies.The P max degradation in HJ cells is driven by significant decreases in fill factor (FF) and open-circuit voltage (V oc ), whereas in the PERT cell technology, degradation primarily results from short-circuit current (I sc ) and V oc losses, as shown in Figure 4. IBC cells exhibited a slight increase in power, despite a decrease in both I sc and V oc .F I G U R E 2 Change in P max for all replicates of each cell type along with a 95% confidence interval in the measurement at each read point, up to 2000 h of UV irradiation under front-side exposure.A negative value indicates degradation.Most of the cell types underwent linear power degradation F I G U R E 3 Final change in I-V parameters of test cells after 2000 h of UV exposure testing, partitioned by cell technology and cell type.The data corresponding to rear-side exposure of bifacial cells is indexed as +.The same side of the cell was illuminated during the I-V measurements as it was facing the UV lamps in the screen test.This shows the UV sensitivity of the front and rear sides of the bifacial cell.As a guide to the eye, a 0% change is indicated by a red dotted line.The 25%-75% confidence intervals are denoted by boxes, and the mean and median are indicated by a square and crossbar, respectively F I G U R E 4 I-V curves for bifacial (A) A HJ and (B) D PERT cells, obtained with front-side illumination (blue) and rear-side illumination (red), before (solid) and after (dashed) 2000 h of UV exposure testing on the same side 3.2 | UVID and Si/passivation damage displays the SIMS depth profiles of Si, O, and H elements acquired on the ITO-etched surface of the A HJ sample.The depth scale was chosen so that the etched surface would correspond to the ITO/(p)a-Si interface.It should be noted that only the H concentration is quantified and presented in atomsÁcm À3 (right y-axis), whereas the intensities of other elements are presented in arbitrary units (left y-axis).There are two inflections in the H profile (red), which correspond to the a-Si and c-Si regions, respectively.The H profile of the UV-aged sample shows higher H concentration across F I G U R E 5 (A-C) SIMS depth profiles of unaged and 2000-h UV-aged samples of (A) ITO-etched A HJ , (B) Si x N y -etched B IBC , and (C) Si x N y -etched D PERT .(D) Spectral reflectance of test cell front surfaces, before (solid) and after (dashed) 2000 h of UV exposure testing: A HJ (red), B IBC (orange), and D PERT (green) the (p)a-Si and (i)a-Si layers and lower H concentration near the a-Si: H/c-Si interface and c-Si bulk.The loss of H near the a-Si:H/c-Si interface would indicate that there is less H available for passivation of dangling bonds or interfacial defect density, increasing carrier recombination and eventually decreasing V oc .This is also consistent with the I-V data for the HJ cell in Figure 4. Manzoor et al. showed that passivation of the c-Si surface with an intrinsic a-Si:H layer degrades over time under illumination.49This degradation is mostly due to an increase in interface defect density, which induces poor chemical passivation.Cho et al. also reported that HJ solar cells with a low H content have a poor a-Si:H/c-Si interface.50The deterioration of the interface is attributed to (i) voids created by insufficiently passivated c-Si surface dangling bonds, (ii) voids formed by SiH 2 clusters, and (iii) Si particles caused by gas phase particle formation in silane plasma.Another potential mechanism for UVID is chemical changes to the passivation layers, such as oxidation or stoichiometry.XPS is sensitive to changes in surface chemistry (chemical bond environment) and complements SIMS.To verify the chemical degradation mechanism, we acquired XPS depth profiles on the front surface of the cell at every 1 min of sputtering.The XPS spectra from the In 3d, O 1s, and Si 2p for A HJ are shown in FigureS5.Only subtle differences were observed between the unaged and aged samples, suggesting that there are no major changes in the chemistry or bond configuration at the cell interfaces beyond the H measured by SIMS.
Figure 4B.In order to examine the underlying degradation mechanism, other characterizations like SIMS, XPS, and optical reflectance measurements were performed on this cell type.The SIMS profiles of the D PERT cells are shown in Figure 5C.Like in the IBC cells, the H profiles (red) show an increased H concentration at or near the Si x N y /SiO 2 interface and underneath the Si bulk after UV aging.However, there are some changes in the Si, O, and N profiles of the n-PERT cell that could have resulted from UV aging, which is verified by XPS and reflectance measurements.

Figure 6 shows
photoinjected electrons could knock out the protons and subsequently migrate along the interface to depassivate Si-H bonds and increase defect generation.For the IBC cell, this energy could easily be supplied by the electrons injected from Si by UVA-340 lamp exposure.It should be noted that the oxygen availability in an encapsulated module would be less, limiting degradation occurring through oxidation-although any monolayer effects at the passivation/Si interface would require minimal oxygen.
Of further interest is the superior performance and durability of the conventional Al-BSF cells against UVID; these cells exhibited F I G U R E 6 High-resolution XPS spectra obtained on the front surfaces of UV-aged (for 2000 h) and unaged D PERT samples for Si 2p, N 1s, and O 1s.Profiles are shown at three regions: the Si x N y surface, SiO x /Si interface, and Si bulk.The peak intensity is normalized to that of the unaged sample the least degradation in each I-V parameter.The front and rear surfaces of these cells are typically passivated by establishing built-in electric field results from the spatially non-uniform concentration of the dopant atoms in the crystal lattice.This creates a high-low junction (n + p in the front or p p + in the rear) that repels the minority carriers to limit both the surface and bulk recombination losses.We hypothesize that the front surface field and complete blocking of UV irradiation by the Al on the rear helps reduce the UV sensitivity of the surface recombination velocity in Al-BSF cells.Modern highefficiency cell technologies have reduced the surface field associated with higher sheet resistance, for example, with the use of selective emitters and more reliance on passivation based on termination of surface dangling bonds and charges within the dielectric layers that can achieve reduced recombination currents.[56][57][58]Cell types with these surface passivation destabilized by UV irradiation do not have the benefit of as large an extent of surface field for resistance to UVID.

Figure 7
shows the change in I-V parameters for cells illuminated under open-circuit and short-circuit conditions.The majority of short-circuited cells (shown in red) exhibited greater power loss than the open-circuited cells (shown in black).Uniquely, the P max loss of A HJ cells was greater under the short-circuit condition F I G U R E 7 Change in I-V parameters of the test cells after 2000 h of UV exposure on the front side, color-coded by electrical load configuration (black = open-circuit; red = short-circuit).The I-V measurements were done by illuminating the front sides of the cells.As a guide to the eye, a 0% change is indicated by a red dotted line (À18%) than under the open-circuit condition (À9%), which was caused by the combined losses in I sc (À1.7%),V oc (À3.1%), and FF (À14%).Another cell type, C PERT , showed a power decrease of À7%; by contrast, the change in E PERT cells was insignificant (less than 1%).On the other hand, G PERC and H PERC cells exhibited P max degradation of À4% and À1%, respectively, suggesting that PERC cells are less susceptible to hot-carrier damage.The data indicates that the hotcarrier injection can also occur under the open-circuit condition, but more concentrated sunlight is required to cause a similar effect.Here, we have identified different degradation mechanisms based on representative examples of different silicon PV cell technologies; however, their occurrence depends on the details of the cell make, including manufacturing processing (time, temperature, and other parameters), dopant (concentration and depth profiles), and thin-film layers (including their thickness and the use of interlayers).No single complete degradation model is developed here; instead, this study indicates that multiple models are required and further clarifies the enabling mechanisms (loss of hydrogen, oxidation, and hot-carrier damage).

Figure 9 (
Figure 9 (bottom); the source spectra of the UVA-340 fluorescent lamp used in our UV test experiment and the spectra of the sun (AM1.5G,from IEC 60904-3) for the spectral range of 250-500 nm are shown in Figure 9 (top).We also assessed the benefits of UVblocking encapsulants in mitigating the impact of UVID and increasing cell performance.For analytical modeling, we considered the optical response of textured Solite glass (coated with a porous silica AR g ), UV-transparent and UV-blocking POE encapsulants, AR c solely composed of Si x N y , and a silicon solar cell.Only the initial performance of the packaging materials is considered in the model and the effects of field and passivation schemes used on the front versus rear sides.As the I-V results show, different cell technologies underwent different degradation mechanisms.Power loss in HJ cells was dominated by a significant drop in V oc and FF, whereas IBC, n-PERT, and p-PERC cell technologies exhibited greater I sc and V oc losses.SIMS characterization showed that UVID is related to hydrogen redistribution in the cell.After UV exposure, higher H concentration was seen near the passivation/Si interface and Si bulk, where bond breaking of hydrogen that passivated the cell surface led to excess H atom clusters that facilitated carrier recombination and deteriorated the cell performance.A separate degradation mechanism (e.g., oxidation at Si interface) was identified for one of the n-PERT cells using XPS and optical reflectance characterization.UV exposure under a shortcircuit configuration of a cell seems to induce greater power degradation than an open-circuit configuration, suggesting a third mechanism related to the hot-carrier injection.The optical modeling of spectral interactions at c-Si module layers and their interfaces demonstrated attenuation of UV radiation by absorptance in the Si x N y bulk and reflectance by the encapsulant/ Si x N y /Si stack.The optical performance of the AR c makes UVID further sensitive to the cell construction, as its composition, density, and thickness affect the damage susceptibility and rate of degradation.For our experiment, we determined an acceleration factor of 2.5Â based on the intensity and possible reciprocity of the UV lamp, and an additional acceleration of 2Â between the bare cell specimens examined in this study and a fielded PV module.Using this correlation, we estimated average and maximum power degradation rates of À0.12%Á year À1 and À0.73%Áyear À1 , respectively, for a PV module containing UV-transmitting encapsulant.Although we demonstrated that UVID reduces the overall power output of modern cell technologies, including higher degradation rates on the back of bifacial cells, these effects may be mitigated by tailoring the UV-blocking characteristics of new encapsulant formulations, using spectral-specific antireflective coatings, and developing UV-stable PV cell designs.

1
Details of cell technologies considered for identical UV screen tests, depending on the sample type, at the National Renewable Energy Laboratory (NREL) and SLAC National Accelerator Laboratory (SLAC) a National Renewable Energy Laboratory (NREL).b SLAC National Accelerator Laboratory (SLAC).