Assessing the stability of p + and n + polysilicon passivating contacts with various capping layers on p-type wafers

Polysilicon (poly-Si)-on-oxide passivating contact structures (POLO/TOPCon) enable high-efficiency solar cells as they simultaneously provide a very high level of surface passivation and a high conductance for either electrons or holes. The ease of incorporation with existing manufacturing lines and their tolerance for high-temperature processing has increased the wide acceptance of this structure in the PV industry. In this report, we explore the effects of short high-temperature annealing required for effective hydrogenation and formation of ohmic screen-printed contacts across a wide temperature range (636 ◦ C – 846 ◦ C) on the stability of passivating contact structures. We study this on p-type c-Si substrates with phosphorus-doped (n-type) or boron-doped (p-type) polysilicon contacts capped with either an AlO x or SiN x coating. Our experimental results show that irrespective of the poly-Si doping type, AlO x -capped samples suffer a loss in surface passivation across the investigated temperature range, while SiN x -capped samples show an improvement at lower annealing temperatures. Above 744 ◦ C, severely ruptured blisters occur for the samples coated with a SiN x layer, leading to lift-off of the poly layer in extreme cases, and in all cases, significant surface passivation losses, up to 99%. A study of the long-term stability of these fired samples under 1-sun illumination @ 140 ◦ C shows that they suffer from both bulk and surface-like instabilities. Two degradation cycles were observed: the first, a boron-oxygen light-induced degradation (BO-LID) observed after 5 min, with capture cross-section ratios of 15.8 – 19.2, and a slower secondary degradation, similar to light and elevated temperature-induced degradation (LeTID), with maximum degradation reached after ~ 14 days. The presence of a silicon nitride layer does not appear to influence the kinetics of post-degradation recovery. Our results suggest that the effect of firing may be influenced by the polarity of the bulk c-Si or perhaps the chemistry of the SiN x film and highlight that passivating contact structures based on p-type c-Si may offer better long-term stability than those based on n-type c-Si.


Introduction
The solar cell market is currently dominated by silicon solar cells, specifically, the passivated emitter and rear cell (PERC) architecture [1]. Efficiency limits imposed by high recombination losses in the phosphorus-doped emitter of the PERC cell have prompted researchers to develop alternatives. One of the most promising solutions is using passivating contacts based on SiO 2 /polysilicon (SiO 2 /poly-Si), in which the most popular approach, currently, is a cell architecture based on an n-type substrate with a boron emitter on the front with a rear SiO x /poly-Si stack. Remarkable surface passivation and extremely low recombination current density parameter, J 0 , less than 5 fA/cm [2,3,], have been shown, with demonstrated record efficiency of 26.1% on an industrial n-type wafer design. While on p-type Si, Haase et al. utilized a rear contact geometry to achieve a record efficiency of 26.1% [4], and Richter et al. demonstrated a both-sides-contacted cell with 26% efficiency [5].
However, recent studies have raised concerns about the stability of passivating contact structures, highlighting potential surface and bulk instabilities resulting from the high-temperature firing step required for metallization. Winter et al. observed bulk-related BO defects in p-type Si and surface degradation and recovery on both n-and p-type Si after light soaking fired samples with n-poly contacts. It was observed that the changes in surface passivation also occurred during dark annealing, with J 0 increasing up to a factor of five after light soaking [6]. Kang et al. also showed that similar surface-related degradation and recovery occurred in n-type Si with poly contacts, deposited either via plasma-enhanced chemical vapor deposition (PECVD) or low-pressure chemical vapor deposition (LPCVD). In that study, it was also observed that the rate of surface degradation was influenced by the temperature during light soaking (LS) and that the silicon nitride layer after firing was required for recovery of surface passivation [7]. Chen et al. showed that in n-type Si, a higher LS temperature accelerated the degradation and recovery kinetics and reduced the extent of degradation and suggested that a complete degradation and recovery cycle was sufficient to mitigate the responsible defect [8].
Although the high-efficiency potential of p-type poly has been demonstrated in the lab with a 26.1% efficient POLO-IBC structure [9], the firing and long-term stability of p-poly structures have not been extensively studied. On p-type Si with poly contacts, Hollemann et al. observed a J 0 increase after firing above 600 • C and highlighted that the changes in J 0 were influenced by the speed of the belt furnace [10]. Also, in a previous communication, we showed the injection-dependent lifetime spectroscopy (IDLS) results of both AlO x and SiN x -capped n-and p-poly samples, tested under accelerated conditions [11]. In that work, we observed a degradation during LS and dark annealing (DA) under accelerated testing conditions for AlO x and SiN x -capped samples. For AlO x capped samples, while p-poly Si samples were stable under the testing conditions, n-poly degraded at a similar rate during LS and DA, with no observable recovery in lifetimes occurring during the testing period. For SiN x -coated samples, the degradation was positively correlated with firing temperature, with higher firing temperatures showing more degradation and complete recovery after 100 s. Such accelerated conditions allow testing to be carried out in timescales in seconds by accelerating the degradation/recovery kinetics [12][13][14][15]. Different reasons have been suggested for the instabilities observed, including an excess of hydrogen in the interface layers [16] and increased thermal stress during the firing process or a combination of them [10].
In this paper, as an extension to previous work [11,17], we investigate the effect of high-temperature firing on p-type Si symmetrically coated with n-and p-poly layers. Further, we examine their long-term stability during 1-sun illuminated annealing conditions. We apply Fourier transform infrared (FTIR) spectroscopy and time-of-flight secondary ion mass spectroscopy (ToF-SIMS) to analyze the hydrogen content in the cell precursors and correlated them to the observed changes in passivation. Specifically, we show that the firing of p-type Si with poly contacts will improve surface passivation under certain conditions. We also show that above a certain temperature threshold, firing is detrimental and can lead to blistering of the poly layer and significant surface degradation. Subsequently, we show that light soaking can lead to bulk and surface instabilities, however, they are significantly less severe than observed in n-type Si. This may imply that p-type Si with passivating contacts may be intrinsically more stable than their n-type counterparts.

Methodology
The test structures studied in this work consisted of p-type (borondoped) Czochralski, 20 Ω-cm Si wafers. All samples were etched in KOH solution to remove saw damage which produced a planar surface and subsequently cleaned using the Radio Company of America (RCA) methodology. A 1.5 ± 0.2 nm thick silicon oxide layer was grown on both sides by a wet chemical process using ozone. The oxide was capped with a 200 nm-thick poly-Si deposited using a CENTROTHERM EUROPE 2000 low-pressure chemical vapor deposition (LPCVD) with in-situ n-or p-type doping using gas precursors PH 3 and BCL 3 respectively. After the poly Si deposition, the samples were annealed for 30 min in a Tempress TS81004 tube furnace at 820 • C. Following this, the surface silicon oxide grown on the poly-Si surface was etched using a 40% HF solution, leading to a poly-Si thickness of about 140 nm. For each doping type (n + or p + ), the samples were split into two groups: one group was symmetrically coated with a 100 nm thick SiN x layer with a refractive index of 2.05 at 600 nm deposited using an inductively coupled plasma chemical vapor deposition (IC-PECVD) Singular XP tool. For the other split, the sample was symmetrically coated with a 10 nm atomic layer deposition (ALD) AlO x layer using an Ultrafast ALD tool by SoLayTEC and subsequently annealed at 425 • C for 30 min in an N 2 atmosphere.
All samples were subsequently fired in an industrial belt-firing furnace (SCHMID/Sierratherm Thermal Furnace System), using a typical temperature profile for the firing of metallization pastes, at a peak (measured) temperature range between 646 • C and 836 • C, and a belt speed of 5.6 m/min. Some samples subsequently had a dielectric etch in HF and followed by a rapid thermal anneal (RTP) in an N 2 atmosphere at 300 • C.
For the stability studies, the fired samples were split into two: one group with a SiN x capping layer and the other group with the SiN x layer etched in dilute HF. Both groups were annealed at 140 • C on a hot plate under ~1-sun illumination from four halogen lamps. The hot plate temperature was verified with a digital multi-meter with K-type thermocouples, and the illumination intensity was calibrated to 1 kWm − 2 using a ThorLabs PM100D power meter. The electronic quality of the structures was evaluated by quasi-steady-state photoconductance decay (QSSPC) measurements using a Sinton WCT-120 lifetime tester. The effective minority carrier lifetime, τ eff was extracted at Δn = 2 × 15/cm 3, and J 0 was extracted using the Kane and Swanson method [18]. Time-of-flight secondary ion mass spectrometer (ToF-SIMS) profiling was done using an ION-TOF TOF-SIMS 5 spectrometer, and Fourier transform infrared spectroscopy (FTIR) was carried out on a Perki-nElmer Spectrum Two FTIR tool. Raman analysis was carried out on a Reinshaw InVia 2 Raman tool.

Impact of firing temperature on n-and p-type poly-Si with various capping layers
First, we investigate the effects of an industrial belt-furnace firing step on the performance of samples with phosphorus-doped and borondoped poly-Si/SiO x layers. For each case, firing at a peak temperature, T F , between 646 • C and 836 • C was carried out on samples with either an AlO x or SiN x capping layer. Before firing, the p-poly and n-poly samples have quite different τ eff and J 0 values. Most p-poly samples had τ eff values around the 1 ms mark and J 0 values above 70 fA/cm 2 . The n-poly samples had τ eff values greater than 3 ms and J 0 values lower than 20 fA/cm 2 . Within each category, the differences in τ eff resulted from spatial non-uniformity across wafers, most likely resulting from experimental non-uniformities in poly-Si deposition or dielectric deposition.

AlO X capping layer
For both p-and n-poly samples, firing resulted in an increase in J 0 and a decrease in τ eff, for all investigated peak firing temperatures. For the p-poly samples, firing above 725 • C led to a noticeable increase in J 0, which was even more significant after firing at 836 • C. Overall, the firing temperature did not have a strong correlation with J 0 , with an average J 0 increase of 32 fA/cm 2 observed for all investigated firing temperatures. This remarkable stability in surface passivation provided by the ppoly AlO x -capped layers is also evidenced in the modest decrease in iV OC , with an average loss of ~9 mV across all firing temperatures.
For the n-poly samples, the reduction in τ eff increased with increasing peak firing temperature. On average, τ eff decreased by ~20% across all firing temperatures for T F between 675 • C and 725 • C, with a greater decrease for T F > 725 • C and even up to 60% loss after firing at 836 • C.
The changes in τ eff were related to a reduction in surface passivation, with a notable increase in J 0 with increasing firing temperature. A maximum increase of more than 400% was observed for the samples fired at 836 • C. Such severe surface degradation observed in n-poly samples at high temperatures can significantly impact device performance, as seen in the changes in iV OC , with losses of up to 35 mV. On the other hand, firing at a lower temperature of 675 • C caused a slight decrease of ~2 mV. From Fig. 1, we observe the superior firing stability of p-poly samples over their n-poly counterparts, which reduces the comparative advantage of n-poly samples after a high-temperature firing step, as illustrated by closer iV OC values to p-poly samples after firing.

SiN X capping layer
In Fig. 2, the τ eff , J 0, and iV OC for double-sided SiN x -capped p-and npoly samples, before and after firing are presented. Like for the AlO xcapped samples, the starting τ eff is greater for n-poly samples, illustrating their superior passivation compared to p-poly samples. Except for T F ≥ 792 • C, firing improved the level of surface passivation of both p-and n-poly samples, as evidenced by the higher (lower) values of τ eff (J 0 ) after firing. In p-poly, the average minimum improvement in τ eff (+120%), J 0 (-58%), and iV OC (3%) occur after firing at the lowest temperature of 675 • C. Increasing T F led to even more significant improvements, although a higher temperature gradient at higher temperatures may have caused increased sample-to-sample variability.
Likewise, for the n-poly samples, enhancements in passivation quality were also observed after firing. However, the improvements after firing do not correlate with T F . Interestingly, although the starting iV OC values were higher in n + poly than in p + poly samples, the improvements after firing were at par, with even less variability among the p-poly samples. These observations are in stark contradiction to our previous work [8], where we showed that firing at similar T F ranges as used in this work was detrimental to n-poly passivation on n-type c-Si. Kang et al. also reported that firing at 700 • C led to a deterioration of surface passivation for both p-and n-poly on n-Si, with greater stability observed in p-poly samples [19]. The major difference between this study and those was the polarity of the Si bulk (p-type in this study versus n-type in the others). It is thus possible that the stability of the surface passivation after firing is dependent on the polarity of the bulk material and requires further investigation.
For both polysilicon doping categories (p-and n-poly), firing at T F ≥ 792 • C was detrimental and caused a severe loss in surface passivation. The effect of firing-induced damage was more evident in p-poly samples fired at 836 • C, with over 97% loss in τ eff and J 0 increasing by over two orders of magnitude (>1000 fA/cm 2 ) and a significant reduction in iV OC (up to 132 mV). For the n-poly samples, the deterioration, although significant, was to a lesser extent than for the p-poly samples. For T F ≥ 792 • C, blistering of the SiN x /poly layers was seen in both p-and n-poly samples and is likely responsible for the disparate firing response at these higher temperatures.  solution. Fig. 3a and b shows images taken from an optical microscope and SEM, respectively, before firing. After firing at 836 • C, ruptured blisters of the SiN x film can be seen (white patches, Fig. 3c), which were not seen before firing (Fig. 3a). Cross-sectional SEM imaging ( Fig. 3d) reveals delamination of the SiN x /poly layers in some of the samples on which blistering is observed, indicating a physical removal of the passivating contacts on those samples. The blistering is likely caused by an infusion of excess hydrogen, released from the SiN x layers during firing. Blister formation is suggested to be due to an accumulation of hydrogen at the poly/SiO x /c-Si interface, with more significant blistering observed in PECVD-deposited a-Si films, due to the hydrogen-rich precursor gases and the deposition conditions employed [20,21]. Such excess hydrogen may build up in the poly-Si interface and eventually lead to delamination of the poly layer [20][21][22], leading to a significant loss of surface passivation as seen in Fig. 2. To further investigate the role of SiN x in the formation of blisters, some p-poly samples with SiN x capping layers were dipped in HF solution to strip off the hydrogen-rich SiN x films before firing at 836 • C. The samples were completely hydrophobic to ensure the complete removal of the SiN x films before firing. Fig. 3e and f shows the micrographs of the samples after the HF dip and firing, respectively. The ruptured blisters seen in SiN x -capped samples ( Fig. 3c) after firing were not visible in samples in which the nitride layers were stripped prior to firing and provide strong evidence to suggest that the presence of the SiN x layer during firing is a prerequisite for the blistering to occur. Besides providing a rich source of hydrogen during firing, the layer will also act as an effusion barrier to prevent the escape of already released hydrogen. We note, however, that, unlike SiN x -capped samples in which the lifetimes improve after firing at T F < 792 • C, the firing of the uncapped samples at 836 • C still led to a loss in surface passivation. This result implies that blistering and delamination of the SiN x /poly layers is likely not the only factor contributing to the deterioration in surface passivation after firing at higher temperatures, which was shown before [17,23], but requires further investigation to clarify the cause. Noteworthy is the fact that the measured τ eff (J 0 ) after firing was one order of magnitude higher (lower) compared to the samples which were capped with the SiN x layer during firing.
In finished cells, ruptured blisters can form a pathway for metal diffusion into the material and potentially allow metals to come in direct contact with Si during firing and exacerbate device degradation [21]. There have been suggestions that blistering and lamination are more severe in samples with chemically-grown tunneling oxide layers due to their off-stoichiometry and lower density [21]. However, this assertion disagrees with recent work by Firat et al. where more blistering occurred in devices that featured thermally-grown tunneling oxides [22]. Fig. 4a displays the Raman spectra of n-(red) and p-(blue) poly films before (dashed lines) and after firing (solid lines) at T F = 836 • C. As a reference, the Raman spectrum of a-Si on a glass slide (transparent in Raman spectroscopy), scaled to have a similar peak area to those of the poly samples, is also shown (green). The clear difference in the peak position of the reference a-Si (475 cm − 1 ) vs poly-Si (515-520 cm − 1 ) and c-Si (520 cm − 1 ) suggests that the poly layers were not in their hydrogenrich a-Si state, but almost fully crystallized before firing. Therefore, the blistering during the firing process was likely caused by hydrogen from the SiN x capping layer. The reduction in the full width half maximum (FWHM) of the characteristic c-Si peak during the firing process can be explained by the relaxation of stress in the microstructure of the poly-Si layer, agreeing with the expectation that firing may contribute to further crystallinity, as observed on the p-poly sample. As a result, the noncrystallinity of the poly-Si layer can be ruled out when considering the cause(s) of lifetime degradation after firing. Fig. 4b shows a clear trend of decreasing N-H stretching absorption band peak area extracted from Fourier transform infrared (FTIR) measurements after firing at different temperatures. The lower N-H bond density after firing suggests a higher release of hydrogen from the SiN x layers at higher temperatures, potentially leading to a higher accumulation at the poly/SiO x /c-Si interfaces. However, we should state that the samples used in this work featured planar surfaces, which have been suggested to have more significant blistering and delamination than textured surfaces due to a lower film adhesion force [20,24]. Notwithstanding, the work by Firat et al. carried out on textured surfaces also exhibited blistering. The importance of the blistering and delamination effects requires more attention, especially given that the temperature range in which this is typically observed coincides with the high-temperature firing sequence needed for the metallization of solar cells.
The lower hydrogen content in LPCVD Si films, higher deposition temperatures employed and the already high crystallinity (Fig. 4a) of the passivating contact layers before firing make the formation of blisters in this work, unexpected. Notably, our work shows that at T F ≥ 792 • C, Bdoped poly samples exhibit higher blistering (and post-firing degradation) than P-doped poly samples. This again is in contradiction to enhanced post-firing stability observed in p-poly samples in other works [19,25]. A notable difference in the structure used in this work and those in other works, including from the authors of this work, is the base doping polarity of the c-Si wafer. In other works, firing-induced degradation occurs when SiN x is used as a capping layer (irrespective of the poly doping) in n-type c-Si, a significant improvement is observed herein in p-type c-Si. Thus, suggesting that the firing response in poly contacts is dependent on the doping polarity of the Si bulk. This may be related to the different charge states of hydrogen in the c-Si bulk, which depends on the Fermi level, and the resulting diffusion coefficient, particularly during the heating and cooling steps. There is also the possibility that the specific deposition tool employed (IC-PECVD) or the composition of the nitride film may play some role in the observed behaviour. In Fig. 5, the percentage change in J 0 after firing polysilicon contacts with capped SiN x films is shown, extracted from other works in the literature. The symbols in red are for n-type wafers, while those in blue are for p-type. For most of the studies examined, various processing parameters such as base doping, SiN x film deposition parameters, and firing belt speed were varied. We observe that most of the J 0 increase after firing occurs in n-type c-Si samples. A notable exception is the work by Morisset et al. on n-type c-Si where reductions in J 0 after firing were observed for samples on which deposition parameters were optimized for a lower hydrogen content [55]. On the other hand, on most p-type c-Si substrates, there was a net reduction in J 0 after firing. We note, however, that there are limited studies on passivating contacts on p-type c-Si. Nonetheless, our results indicate that the base doping polarity or perhaps the composition of SiN x or deposition tool employed, may influence changes in the surface passivation response to firing.
Regarding field-effect passivation, dielectric films such as AlO x and SiN x deposited on silicon provide fixed charges which can modify the surface state of materials, with the former exhibiting negative fixed charges and the latter positive charges [26,27]. However, the Debye length in the grown poly-Si sample is usually much smaller than the thickness employed in making passivating contacts. Thus, the influence of the fixed charges by dielectrics is screened by the highly doped poly layer, with exceptions in very thin poly-Si with low doping. In a separate firing study (not shown), the authors did not observe any significant difference in the diffused dopant concentration after firing. Hence, firing was not expected to influence the concentration of electrically active dopants within the poly-Si layer/c-Si substrate. The AlO x samples employed in this study were annealed at 425 • C for 30 min in N 2 , before the firing results are shown in Fig. 1.

Dehydrogenation after HT firing
As observed in the previous sections, the passivation quality of fired samples is strongly related to the release of hydrogen and its complicated interaction with different layers of the structures. Indeed, while the positive effects of hydrogenation in improving the performance of solar cells are widely acknowledged [28,29], hydrogen has also been shown to be responsible for the formation of recombination centres [30], dopant deactivation [30], formation of structural defects [31] and the creation of vacancy-hydrogen complexes [32], leading to severe bulk and surface lifetime degradation [33,34].
To further understand the role of hydrogen, samples fired at 846 • C were further annealed in N 2 at 300 • C. It has been previously reported that short dehydrogenation in N 2 after a high-temperature firing could be beneficial, since an excess of hydrogen released during firing may be responsible for loss in surface passivation. Kang et al. demonstrated that a short dehydrogenation anneal was sufficient to reverse degradation observed after firing [19]. Two groups were presented: one with the capping layers (AlO x and SiN x ) present and the other with the capping layers etched off in HF, prior to annealing. The changes in J 0 after annealing in N 2 are depicted in Fig. 6, and for reference, the J 0 values before and after firing are also presented. For all groups, annealing in N 2 after HT firing did not lead to improvements in surface passivation. In samples where J 0 reduces or stabilizes with continual annealing, the level of surface passivation was still considerably lower than in the pre-fired state. On the AlO x (and AlO x -stripped) samples in Fig. 6 a and b, annealing at 300 • C in N 2 atmosphere did not lead to improvements in   [8,17,[53][54][55]. Blue symbols are for p-type c-Si, and red for n-type c-Si. surface passivation. On the contrary, the J 0 values increased after annealing, except in the n-poly sample without an AlO x layer, which remained stable during the annealing. The increase in J 0 was greater for the n-poly samples (which had lower initial J 0 values compared to p-poly). The reduction in surface passivation is likely due to a loss in hydrogen after firing, as observed by Van de Loo et al. It was noted that the removal of AlO x capping layer led to a significant reduction in hydrogen concentration and that samples with an AlO x capping layer showed much greater stability during annealing in N 2 [35]. AlO x typically contains 2-3% at. % H which provides chemical passivation for defect passivation. However, at about 400 • C, effusion of molecular hydrogen takes place, and the ability of the AlO x layer to act as a good effusion barrier reduces [36]. Before annealing, the samples had undergone a high-temperature firing process (846 • C) and had already lost a significant amount of hydrogen, which explains the loss in surface passivation after firing. Annealing in N 2 would further dehydrogenate the samples and lead to increased surface deterioration, as observed. The changes in surface passivation, as indicated by the evolution in J 0, may imply that a significant amount of hydrogen was already released into the bulk or effused during the HT firing process and annealing in N 2 was causing a further de-hydrogenation of the samples. For n-poly, the J 0 of the uncapped layer (53 fA/cm 2 ) was much lower than the capped layer (122 fA/cm 2 ) after de-hydrogenation, although they had similar J 0 values before the dehydrogenation anneal. Removing the AlO x layer makes the poly-Si stable during further dehydrogenation. It is possible that with further annealing, the remaining H contained in the AlO x layer is continually released and builds up at the poly interface since the annealing temperature is not high enough for quick effusion. This explanation does not hold for p-type poly, where the uncapped samples show de-passivation with N 2 annealing. This disparate response to de-hydrogenation in n-poly is not clear and requires further examination.
On the SiN x samples, annealing without a nitride capping layer was more detrimental to surface passivation for both p-and n-type poly. After 15 min of annealing, samples with a SiN x capping layer, although with ruptured blisters (Fig. 3c) had considerably lower J 0 (more than 1 order of magnitude) than samples w/o a capping layer. This is in contrast to observations by Kang et al. [23], where it was shown that annealing in N 2 after an 800 • C firing step improved surface passivation. Kang   been removed, annealing for up to 40 min at 300 • C led to improvements in surface passivation because of reduced H concentration around the tunneling oxide layer. Presumably, however, these samples were not blistered, which is not stated in the paper. Without a capping layer, H effusion is expedited, leaving the precursors with much less H for passivation. There is also the effect of reduced light-coupling due to SiN x removal, which could also negatively impact the device.
It is clear that for both AlO x and SiN x -coated samples, the removal of "excess" hydrogen via a dehydrogenation step can further deteriorate surface passivation. Although in the blistered samples, the nonuniformity caused by local removal of the poly contacts in some regions, makes a direct comparison with non-blistered samples difficult. But the results show that a dehydrogenation step is not necessarily a good step and may even lead to further surface degradation.

Effect of long-term stability testing on p-and n-poly layers
In this section, we explore the effects of long-term stability testing for SiN x -coated poly samples, under 1-sun illumination conditions at 140 • C, following the fast-firing step.
Herein, we examine n-and p-poly structures, capped with SiN x films.
In Fig. 7, the lifetime τ eff (and corresponding J 0 ) extracted at an injection level of Δn = 2 × 10 15 /cm 3 as a function of cumulative light soaking time, is depicted for p-poly (a, b) and n-poly (c, d). Two degradation cycles were observed for all samples fired between 646 and 744 • C. We recall from the previous sections that the samples with T F ≥ 792 • C suffered severe degradation with ruptured blisters on the samples and had comparatively lower lifetimes at the start of the 1-sun stability test. Two consecutive degradation and recovery cycles, and the onset of a third degradation, were observed. The initial degradation occurred only after 0.08 h in both p-and n-poly samples and was fully recovered after 1 h of exposure. This very fast degradation can be attributed to the wellknown boron-oxygen (BO) [37] defect. The relatively flat J 0 values and increasing iV OC (not shown) during this initial lifetime degradation suggest that the BO bulk defect is likely responsible for this initial degradation which is consistent with other findings [38]. Additional IDLS analysis was carried out on the measured lifetimes of the p-poly samples i) after firing and ii) at the point of maximum degradation after 0.08 h. Using a mid-gap energy level, as the exact energy level remains unknown, capture cross-section ratios, k LS , of the defect formed after light soaking for 0.08 h was evaluated for the samples fired at 700, 725, and 744 • C and was found as 19.2, 15.8, 18 respectively. These values are similar to those previously derived for BO degradation [39].
The degradation was more severe in n-poly samples which featured comparatively higher lifetimes than p-type, with some increase in J 0 observed for higher firing temperatures, and the extent of degradation increased with increasing firing temperature. The BO defect was not observed for T F < 700 • C. The maximum subsequent secondary degradation occurred after ~350 h and exhibited less severity in n-poly than the initial BO degradation, where the lifetime at maximum secondary degradation was still higher than the initial values. In p-poly samples, however, the maximum degradation was more severe than the initial BO degradation and even more so with decreasing firing temperature. The p-poly sample fired at 646 • C reached a maximum degradation after ~800 h. We note that the almost imperceptible changes in the sample fired at 646 • C during the first degradation cycle were not due to the lower starting lifetime, as the same sample shows significant lifetime degradation during the second cycle. The evolution of J 0 , which increases by up to 100% in p-poly and 70% in n-poly, indicates that the changes are primarily due to degradation in surface passivation. Continual exposure to illumination mostly recovers the n-poly samples, but not the p-poly samples, to post-BO deactivation levels. After light soaking for over 350 days, the onset of a tertiary degradation was seen. In n-poly, the increase in J 0 is more significant than the previous degradations observed but less so in p-poly. The high number of handling steps caused some handling marks to appear (PL images not shown), which invariably would contribute to the loss in passivation exhibited. Nonetheless, as Fig. 7 shows, the unfired samples, which were processed and handled similarly, show remarkable stability throughout the illumination duration in both p-and n-poly samples and highlight the fact that the degradation exhibited is firing-induced and occurs irrespective of poly doping.
Next, we investigate the changes in the samples fired at T F ≥ 792 • C after light soaking. As mentioned earlier, these samples exhibited severe blistering, with ruptured blisters and partial delamination of the poly layer observed in some samples (Fig. 3). The decreasing concentration of N-H peaks with increasing firing temperature (Fig. 3b) further suggests a greater release of H due to higher thermal energy. Whilst all samples did not show the same level of blistering and delamination, severe degradation of a lifetime was observed for all SiN x -capped samples fired Fig. 7. (a) τ eff (p-poly) and (b) J 0s (p-poly), and similarly (c) τ eff (n-poly) and (d) J 0s (n-poly) during illuminated annealing at 1-sun and 140 • C. τ eff was extracted at Δn = 2 × 10 15 /cm 3 . NF refers to "non-fired" samples.
at this temperature range. The initial 20 s exposure to light, while leading to a rapid deterioration of almost 50% of the already degraded post-fired lifetime in the n-poly samples, enhances the lifetime in p-poly samples. One reason could be that the p-poly already suffered the most severe degradation with lifetimes <10 μs, making further degradation unnoticeable. Further exposure leads to an unexpected and significant improvement in surface passivation in p-poly samples. The 836 • C-fired samples, which had the most severe firing-induced degradation, improved almost 30-fold (similar changes in the 892 • C-fired sample) compared to the post-firing state, with lifetime enhancement from 4.7 μs to 135 μs, a strong reduction in J 0 from >10,000 fA/cm 2 to 534 fA/cm 2 and an iV OC improvement of 84 mV (not shown). In the n-poly samples, no changes were observed for the 836 • C-fired samples, but improvements were seen in the 792 • C-fired samples. The strong recovery observed, considering the blistering and loss of passivation layers on the poly surface, was quite unexpected. Although the recovery did not lead to lifetime values measured prior to firing within the time frame of exposure (365 days), it does not appear to have plateaued and may continue to improve with further illuminated annealing.
Historically, hydrogen has been shown to contribute to BO-LID and LeTID defect reactivation. In BO-LID, hydrogen dissociation from passivated BO complex at high temperatures (in the dark) allows the recombination active defect to reform upon exposure to light [40]. In LeTID, hydrogen released from a bulk hydrogen reservoir triggers a recombination-active complex [41]. In a separate work on n-type Si wafers, the degradation and complete recovery occurred after ~270 h [8]. The degradation and recovery observed in this study, and an onset of a tertiary degradation after over 365 days, raises serious concerns about the long-term stability of passivating contact structures on p-Si wafers and shows similarities with H-related degradation and recovery earlier referenced.
To investigate the potential role of hydrogen in the degradation and recovery observed after firing, we perform some ToF-SIMS measurements to evaluate the depth profile of H in the samples at various stages of the experiment: prior to HT firing, after firing (at 646 • C and 836 • C), after LS (partial recovery). The samples used for ToF-SIMS analysis were obtained from the same batch as previous sections but fired separately. Again, as noted in Fig. 2, firing at 646 • C led to an improved effective lifetime (752 μs-1629 μs), while firing at 836 • C degraded the samples (>700 μs-~4 μs). Light soaking of the degraded sample (for 17 days) led to significant improvement in surface passivation, highlighting the repeatability and agreeing with the results earlier presented (Fig. 5). ToF-SIMS analysis was then carried out on all four samples to observe the distribution of hydrogen up to 400 nm etch depth.
In the ToF-SIMS result shown in Fig. 8, expectedly, a higher H content is observed in the H-rich SiN x regions. A lower [H] is measured on samples fired at 836 • C for those regions, in agreement with FTIR results earlier presented, showing reduced N-H stretching bond density after the high-temperature firing. One significant difference between the 836 • C-fired samples, which showed severe lifetime deterioration, and the other samples (unfired and fired at 646 • C), is the magnitude of hydrogen build-up at the SiO x interface. The presence of a silicon oxide layer can block H and decrease H flux to underlying layers by orders of magnitude [42]. As such, this is not very surprising and occurs about an order of magnitude higher for samples fired at 836 • C. Further, the concentration of hydrogen in the poly layer and the c-Si is higher in the 836 • C-fired samples. Kang et al. have suggested that such build-up of hydrogen at the interface and excess hydrogenation of the stack is responsible for the deterioration of surface passivation [19]. Our results here agree with that suggestion, as the observed deterioration coincides with the massive build-up of hydrogen at the SiO x interface and an increased hydrogen concentration in the poly layer and c-Si bulk.
Comparison . This suggests that while hydrogen does improve surface passivation, an excess amount is undesirable, in agreement with other findings [17,23,43]. Analysis of the light-soaked sample that showed some recovery also points to similar conclusions, with a higher [H] in the poly-Si/SiN x region than in the non-light-soaked sample but slightly lower in the bulk of the poly and SiO 2 /c-Si interface regions. These ToF-SIMS results aid our understanding of the possible phenomenon underlying the surface passivation recovery after 365 days of light soaking (Fig. 8), but may not be conclusive on their own, as accurately measuring H can be challenging. However, as had been previously suggested, high temperatures may lead to thermal stress at the SiO x /c-Si interface due to thermal coefficient mismatch [10], and a release of an excess amount of hydrogen from the SiN x layer. While hydrogen can help passivate some defects, as already observed, it can also lead to other problems. Thus, the density of defects in Si can be multiplied after a firing step. With the excessive hydrogenation at the interface, in-diffusing H can lead to the formation of platelets by incorporating into strained or dangling bonds at the interface, thus, triggering the formation of blisters [31,44,45]. During the long LS that follows, a gradual diffusion of the H through the bulk and the surface could lead to surface passivation improvements.
The diffusivity of H in Si is a key element of hydrogenation and depends on the charge state of hydrogen [46], temperature, excess carrier concentration [47], and concentration of impurities and defects. Fig. 9 by Chen et al. shows a compilation of hydrogen diffusivity data taken from numerous works and converted into the time taken by hydrogen to diffuse through 100 μm of silicon [48]. Accordingly, for a 130 • C LS process, the time for hydrogen to diffuse through a 100 μm wafer could vary between 1 × 10 5 to 1 × 10 9 s. The cumulative LS time of 365 days is well within this range, as such, makes hydrogen effusion from the interface and Si bulk a likely explanation for the improvements observed. Since the HT firing process may have induced the generation of a large concentration of defects, it may be that hydrogen effusion will take a longer time in these samples, given the reduced diffusivity known to occur in Si as a result of impurity or defect concentrations [49,50]. If the diffusion of excess hydrogen during LS is responsible for the improvement, it is not very clear, however, why the dehydrogenation steps employed, as illustrated in Fig. 6, did not lead to surface passivation improvements. For the blistered samples, it is understandable that the local delamination of the poly Si layers, further deteriorates the surface passivation.
Considering that the degradation and recovery during light soaking are observed only in fired samples, the role of metal complexes formed during the high-temperature process cannot be completely ruled out. In a defect model suggested by Bredemeier et al. metal complexes thought to form following a firing step, can be reconfigured into a recombination-active form during light soaking. Subsequent dissociation yields a highly recombination-active interstitial metal (degradation), which under continued light soaking at elevated temperature, can diffuse to wafer surfaces and crystallographic defects where they become trapped and thus, recombination inactive (recovery). Lesser degradation and a faster recovery in thinner samples have been used to validate this model in high performance mc-Si wafers [51]. The role of metal complexes in the degradation observed in this work, is however left to future work.
The role of the SiN x -capping layer on the stability of the samples during light soaking was also examined. Fig. 10 shows the changes in τ eff , J 0, and iV OC for samples fired and then stripped of SiN x layer prior to LS for 330 h. Similar conditions were used as presented earlier. For samples fired at T F ≤ 744 • C, the results obtained are quite similar for light soaking with nitride layers on. Similar timescales were observed for the first BO-LID degradation and recovery cycle. The onset of the second degradation also occurs in identical timescales to the results presented in Fig. 9. Hydrogen diffusivity data in mono-and multicrystalline silicon converted into the time required to diffuse through 100 μm, compiled from literature by Chen et al. [48]. The area marked in grey corresponds to the light soaking temperature of 130 • C. Fig. 10. Light soaking at 1-sun and 140 • C for samples stripped of SiN x (in HF) after firing, for (a) p-poly: τ eff , J 0s , iV OC ; and (b) n-poly: τ eff , J 0s , iV OC . Note the different scales on the y-axis for n-poly vs p-poly samples. τ eff was extracted at Δn = 2 × 10 15 /cm. 3 . Fig. 7. At higher firing temperatures, (T F ≥ 792 • C), the notable difference is for p-poly samples which do not show the immediate recovery observed with a nitride capping layer. Rather, as seen in the T F = 792 • C sample, a surprisingly higher lifetime is observed after firing, and a strong degradation and recovery kicks in which was not observed on samples that had a nitride layer. The observations from Fig. 10 suggest that the nitride layer neither plays a critical role in affecting the kinetics of the observed degradations and recovery during LS nor is it required for recovery. In addition, we can conclude that the hydrogen involved in the kinetics observed during LS does not originate from the SiN x layer after firing, but from the hydrogen introduced into the bulk and interfaces during firing. Compared to n-type Si, passivating contact structures samples based on p-type Si show remarkable stability during LS. In previous work, we showed that the relative changes in J 0 at the point of maximum degradation during LS for n-type Si with n-poly contacts, compared to J 0 after firing, varied from 166% to 550% [8]. The reader is referred to this work for deeper understanding. In that work, firing was done on the same tool as employed in this work, with similar temperature ranges. The samples, however, were processed differently and on n-type Si wafers, so, direct comparisons may not lead to accurate conclusions. Yet, it is noteworthy that the differences in the magnitude of degradation may suggest that p-type Si may offer better long-term stability compared to n-type. If we consider the n-poly sample fired at 744 • C in this work (which showed the most degradation), at the point of maximum degradation, the J 0 and % increase in J 0 is approximately 19 fA/cm 2 and 73% respectively, while for a similar firing temperature on n-type Si, the values were 62 fA/cm 2 and 480%. The differences are staggering and are worth considering for further investigation in the broader context of increasing market share for n-type-based passivating contact solar cells [1].

Conclusion
In this work, we have studied the changes that occur post-firing in SiN x and AlO x -capped passivating contact samples having either n-or ppoly layers on p-type c-Si wafers. In SiN x -capped layers, below a firing temperature of 792 • C, improvements on the surface passivation occur, while above this temperature, severe degradation up to 99%, occurs. In AlO x samples, firing at all temperatures led to surface degradation. At T F > 792 • C, firing was observed to cause significant blistering and delamination of the poly layer and occurred only in the presence of a nitride capping layer. As such, excess hydrogenation and or trapping of the released hydrogen from the nitride layer during firing, were suggested as leading causes. Subsequent dehydrogenation by annealing in N 2 at 300 • C did not lead to improvements in either SiN x or AlO xcapped samples. The dissimilarity in surface passivation changes after firing, as compared to other works in literature, suggests that the base polarity or the chemistry of the SiN x film plays a role in determining whether passivation improves or degrades after firing.
The results from long-term stability tests carried out under 1-sun illumination at 140 • C, reveal that SiN x -capped samples are prone to surface and bulk-related instabilities. Specifically, BO-LID defects with capture cross-section ratios in the ranges 15.8-19.2 were observed, along with LeTID-like surface defects which occurred after 160 h. The presence of BO-LID defect on high resistivity samples shows the sensitivity of such structures, which might be minimized by utilizing a Gallium-doped substrates. Although Gallium-doped samples have been shown to suffer from other forms of long-term degradation, similar to those observed in this work [52]. The slow degradation observed for the latter is concerning and requires further studies to provide mitigation strategies. A comparison of the extent of degradation in p-type and n-type c-Si, highlights that p-type substrate may provide much better stability. When compared to the literature, the extent of degradation in n-type-based poly-Si samples were observed in some cases to be more than seven (7) times that in p-type and may be related to the variations in hydrogen diffusivity across both doping polarities. Such differences warrant further studies with more direct comparisons on the role of base doping for polysilicon passivating contact samples. Light soaking led to significant surface passivation improvement on blistered samples which were severely degraded after firing. Further analysis of similar samples showed that the recovery may be due to changes in the hydrogen concentration at the SiN x /polySi and around the SiO x interfaces, and further highlights the likely involvement of hydrogen in the long-term stability of these structures. Further, we show that the degradation and recovery kinetics observed during light soaking has a temperature dependence but are independent of a nitride capping layer.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.