Effect of different oxygen precursors on alumina deposited using a spatial atomic layer deposition system for thin-film encapsulation of perovskite solar cells

An atmospheric-pressure spatial atomic layer deposition system operated in atmospheric-pressure spatial chemical vapor deposition conditions is employed to deposit alumina (AlO x ) thin films using trimethylaluminum and different oxidants, including water (H2O), hydrogen peroxide (H2O2), and ozone (O3). The impact of the oxygen precursor on the structural properties of the films and their moisture-barrier performance is investigated. The O3-AlO x films, followed by H2O2-AlO x , exhibit higher refractive indexes, lower concentrations of OH− groups, and lower water-vapor-transmission rates compared to the films deposited using water (H2O-AlO x ). The AlO x films are then rapidly deposited as thin-film-encapsulation layers on perovskite solar cells at 130 °C without damaging the temperature-sensitive perovskite and organic materials. The stability of the p–i–n formamidinium methylammonium lead iodide solar cells under standard ISOS-D-3 testing conditions (65 °C and 85% relative humidity) is significantly enhanced by the encapsulation layers. Specifically, the O3-AlO x and H2O2-AlO x layers result in a six-fold increase in the time required for the cells to degrade to 80% of their original efficiency compared to un-encapsulated cells.


Introduction
Recently, organic-inorganic halide perovskite solar cells (PSCs) have emerged as a promising alternative to existing photovoltaic technologies due to their ease of fabrication and low cost [1].There has been a rapid improvement in device efficiency, reaching over 25%, in only a few years [2]; Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.however, for PSCs to be commercially viable, their instability needs to be overcome, which is most severe at elevated temperatures or under the presence of moisture [3][4][5].Fortunately, full-encapsulation methods have demonstrated encouraging results in protecting PSCs from moisture, even under high humidity conditions, since encapsulation layers can act as a gas barrier, restricting the diffusion of moisture [6,7].Various encapsulation approaches have been proposed to shield PSCs from external stress factors.Among these, thin-film encapsulation (TFE) offers several advantages; TFE can serve as a a compact, pinhole-free barrier against moisture and oxygen while maintaining the solar stack's flexibility for roll-to-roll processing [8,9].Thin films can be applied directly onto PSCs and can be used in conjunction with polymer and glass-to-glass encapsulation [9].
Atomic layer deposition (ALD) is an effective technique to fabricate moisture-barrier layers due to the dense and uniform films it produces [10].However, for the disposition of high-quality TFE using ALD, a high temperature is required, which could damage the perovskite absorber layer [8].Furthermore, conventional ALD suffer from very low deposition rates as well as high cost due to complex and expensive vacuum technology [10][11][12].Therefore, conventional ALD is not typically applicable for low-cost or highthroughput manufacturing.Consequently, atmospheric-pressure spatial ALD (AP-SALD) has gained interest as an alternative approach to overcome these drawbacks [13,14].In contrast to conventional (temporal) ALD in which the chemical precursors are separated in time, in AP-SALD the different precursors are supplied constantly but in different physical locations, while the substrate moves back and forth between the precursor zones at atmospheric pressure.A nitrogen gas shield spatially divides the reaction zones and films are grown by alternatively exposing the substrate to the precursors [15].Therefore, AP-SALD is very suitable for high-throughput manufacturing in roll-to-roll and sheet-tosheet processes, since long chamber purge times are not needed for spatial ALD and the hardware is relatively simple and inexpensive [16,17].Moreover, by adjusting the deposition conditions to allow the precursors to mix in the gas phase, atmospheric-pressure spatial chemical vapor deposition (AP-SCVD) can be performed in the same system, which is attractive due to higher deposition rates [15].
Initial demonstrations of AP-SALD on top of metal-halide perovskite materials have included the deposition of TiO 2 and SnO 2 electron-transport layers on the perovskite of a PSC, as well as a ZnO charge-injection layer on CH 3 NH 3 PbBr 3 in a perovskite LED [18][19][20].However, to date, there has been limited research on TFE layers for PSCs using AP-SALD.
Alumina (Al 2 O 3 ) has a relatively low water vapor transmission rate (WVTR), good adhesion and high mechanical strength, making it a superior TFE layer [21,22].In a previous study, we reported that an AP-SALD system can be used to produce uniform nitrogen-doped AlO x (N-AlO x ) thin films for encapsulation of PSCs [23].The N-AlO x films showed outstanding barrier properties with WVTR values on the order of 10 −4 g m −2 d −1 at accelerated conditions of 65 °C and 85% RH, which was attributed to a high film density, trapping of water molecules by nitrogen defect sites, and a lower concentration of OH-defects on the N-AlO x film surface compared to undoped AlO x films, as water-vapor transmission through thin films can occur by permeation of H 2 O molecules along chains of chemical defect clusters like hydroxyl groups [24].The lowest WVTR was observed for N-AlO x films with 0.28% nitrogen and a p-i-n PSC encapsulated with theses films maintained 80% of its initial power conversion efficiency after exposure to 65 °C and 85% RH for 300 h, whereas an equivalent un-encapsulated device was only stable for 53 h [23].In another recent study, we used an AP-SALD system operated in AP-SCVD conditions to produce uniform zinc aluminum oxide (Zn-AlO x ) thin films for PSC encapsulation [25].The Zn-AlO x films similarly showed WVTR values on the order of 10 −4 g m −2 d −1 at ISOS-D-3 conditions (dark, 65 °C, 85% RH) and the lowest WVTR and largest stability enhancements were observed for films with a Zn/Al ratio of 0.21.A p-i-n PSC encapsulated with this TFE layer maintained 80% of its initial efficiency during 384 h of exposure to the ISOS-D-3 accelerated testing conditions, while an equivalent un-encapsulated device was only stable for 52 h [25].
Thus, these AP-SCVD N-AlO x and Zn-AlO x thin films provide a promising new PSC-encapsulation route, not only for lab-scale studies but also for high-throughput commercialization efforts.However, there remain gaps in our understanding of how alternative deposition conditions, such as the type of oxidant, can influence the barrier properties of TFE.Most often trimethyl aluminum (TMA) is used as the aluminum source and water as the co-reactant for ALD and SALD of alumina, as H 2 O shows facile ligand exchange and has a low environmental impact [26].However, there are several reports that water-based ALD materials suffer from specific drawbacks.For example, oxide films deposited with H 2 O often include oxygen vacancies or hydrogen within their bulk, which can lead to unwanted permeation pathways for moisture [27,28].It has been established that the properties of Al 2 O 3 ALD are strongly affected by the oxidant used, with ozone or an oxygen plasma producing better films than water [29,30].A higher film density and/or lower content of impurities and unreacted OH− groups was observed in the film, due to the higher electrochemical potential and higher volatility of ozone [30].The growth of metal oxides with ozone can also be performed at significantly lower temperatures [31].However, ozone must be generated in situ, which involves additional equipment and cost [32].Hydrogen peroxide (H 2 O 2 ) can be an ideal alternative to ozone or H 2 O, as H 2 O 2 has similar oxidation properties as ozone (oxidation potential of O 3 = 2.1 V versus 1.8 V for H 2 O 2 ) while simultaneously having slightly stronger proton-transfer properties than water (H 2 O 2 pKa = 6.5 versus pKa = 7.0 for water) [33].
The properties of AlO x TFE layers deposited by AP-SALD using oxygen sources other than H 2 O, especially H 2 O 2 , have yet to be characterized in detail and incorporated into PSC device studies.Only a limited number of previous studies have considered H 2 O 2 as an oxidant for conventional ALD [34,35] and to the best of our knowledge, there are currently no published reports on the effect of using H 2 O 2 in a spatial ALD process, which could be partially due to difficulty in dosing pure H 2 O 2 , since it is often supplied as a mixture with H 2 O, hindering analysis of its chemical reaction mechanisms and process conditions [33].
In this study, AlO x thin films are deposited with an AP-SALD system operated in AP-SCVD conditions using H 2 O, H 2 O 2 and O 3 as the oxidant and the deposition rate, microstructure and barrier properties are investigated.The films deposited using different oxidants are then applied as TFE layers in PSCs, which are measured under ISOS testing conditions [36].The stability of the PSCs is significantly enhanced by the TFE layers deposited rapidly in open-air and the different oxidants are found to impact the PSC performance and stability.

Deposition of AlO x thin films with different oxidants
AlO x thin films were deposited in open-air conditions using our custom-built, dual AP-SALD/SCVD system, which employs a close-proximity reactor-head configuration [37,38].We kept all deposition conditions constant and varied only the oxygen source, to gain insights into the effects of different oxygen precursors on the properties of the resulting TFE layers.Trimethyl aluminum (Al 2 (CH 3 ) 6 ) was used as the metal precursor and water (H 2 O), hydrogen peroxide (30% H 2 O 2 in water), or ozone (O 3 in O 2 /N 2 gas) was used as the oxygen precursor.Vapors of TMA, water, and hydrogen peroxide were generated by bubbling inert nitrogen gas through the chemicals.Also, nitrogen was used as a carrier gas to dilute the precursor vapors and deliver them to the AP-SALD/SCVD reactor head, which has parallel gas outlet channels along its bottom surface, as shown in figure 1.
During the study, a constant 25 SCCM nitrogen flow was maintained through the TMA bubbler and 100 SCCM through the water or hydrogen peroxide bubbler.Meanwhile, the flow rates were fixed at 100 SCCM and 150 SCCM through the TMA and oxidant carrier lines, respectively.The metal precursor was delivered to one outlet channel and the oxidant was delivered to two outlet channels (125 SCCM per channel).Based on the vapor pressures of the precursors, their concentrations in nitrogen can be approximated as 0.22 volume % for the metal precursor, 1.2 vol% for the H 2 O, and 0.01 vol% for the H 2 O 2 (with 1.2 vol% H 2 O).For depositions with O 3 , O 2 /N 2 gas (7.5% nitrogen balance oxygen) was supplied to an Atlas 30 UHC ozone generator (Absolute Ozone) set to 80% power and subsequently delivered to the reactor head, again with a flow rate of 125 SCCM per oxygen precursor channel.The expected ozone concentration for these settings is 221.9 grams per normal cubic meter (g Nm −3 ).A 125 SCCM flow of nitrogen was also delivered to each inert gas channel (green channels in figure 1).The deposition conditions are summarized in table 1.Our Previous studies have examined the impact of deposition parameters on the quality of the deposited layer [23,25].In this study, optimized parameters were used to produce pinhole-free, compact and uniform films.Substrates were placed on a heated stage (set to 130 °C) and positioned underneath the reactor head with a small vertical separation distance of 100 μm.The heated stage was oscillated back and forth at the speed of 50 mm s −1 to expose the substrate sequentially to the alternating precursor flows to deposit films with nanoscale thickness control.In this study, experimental conditions were selected such that the inert gas flow and exhaust strength were insufficient to prevent mixing of the reactant vapors in the gas phase during the oscillation of the substrate, directing the system to operate in AP-SCVD mode.The WVTRs of the alumina films were measured using a procedure reported previously [23,25].Briefly, glass substrates were cleaned and a 200 nm Ca film with a 1.8 1.8 cm 2 area was thermally evaporated in a glovebox.The Ca films were opaque with a mirrorlike appearance.The Ca samples were vacuum sealed inside the glovebox and transfer to the AP-SALD system, where the vacuum bag was opened immediately prior to TFE deposition.AlO x films 60 nm thick (verified by ellipsometry) were then deposited directly on the Ca films.The edges of the samples were sealed using epoxy glue in a glovebox then the samples were placed into an ESPEC SH-222 benchtop environmental chamber at 65 °C and 85% RH in the dark (ISOS-D-3 testing condition).A Nikon D7100 camera was used to take pictures of the samples every 15 min to 1 h and IMAGEJ software was used to convert the images to black and white and measure the white areas that corresponded to oxidized Ca film.The WVTR was determined from the rate of film oxidation [39].

p-i-n perovskite solar cell fabrication
PSCs were fabricated using a procedure reported previously [23,25].Briefly, a hole transport layer of PTAA was spincoated onto cleaned FTO substrates and annealed at 100 °C for 10 min in a glove box.The perovskite layer was fabricated by a two-step deposition method.First, to increase the coverage of the perovskite films on the PTAA, the PbI 2 solution was deposited by spin-casting, and then a drop of the mixed solution containing FAI: MAI: MACl was added to the spinning substrate for 30 s.The as-prepared samples were annealed at 150 °C for 8 min, and slowly cooled to room temperature.An electron transport layer of PC 61 BM was spincoated on the perovskite films and annealed at 100 °C for 7.5 min.Then a BCP layer was deposited by spin-coating.To complete the device, 100 nm thick Ag top electrodes were deposited by thermal evaporation.Finally, AlO x TFE layers were deposited directly on the PSCs using the different oxidants.The rapid nature of the AP-SALD/SCVD process, combined with protection afforded by the electron transport layers and Ag electrode, prevent damage to the perovskite film during encapsulation at 130 °C [23].

Characterization
A FilmSense FS-1EX ellipsometer with a wavelength range of 200-1800 nm was used to measure the thicknesses and refractive indexes of the AlO x films, which were modeled using the Cauchy formula.Surface roughness was measured using a Dimension 3100 atomic force microscope (AFM).Scanning electron microscopy (SEM) images were captured using a Zeiss Leo 1530 with an electron high tension of 5 kV.A VG Scientific ESCALAB 250 x-ray photoelectron spectroscopy (XPS) system was used to determine the elements present in the thin films using Al Kα x-rays.The relative intensities of the XPS peaks was of primary interest, such that the C 1s signal was used for calibration, but it is recognized that a more reliable calibration signal should be used in the future if the peak positions are important.
A Sciencetech UHE-NL-150 Class AAA Solar Simulator was used to conduct the J-V measurements under simulated solar illumination (AM 1.5G).The solar cell area was calculated to be 0.1695 cm 2 , an irradiance of 100 mW cm −2 was set using a NIST-calibrated silicon reference solar cell, and the voltage was swept from 1.1 to −0.1 V with a sweep rate of 0.02 V s −1 .To assess the PSC stability, the samples were stored in the SH/SU-222 environmental chamber and their power conversion efficiencies were measured periodically and compared to their initial performance measured at T = 0 h.sample, which is close to the stoichiometric value of 1.5 [40].Replacing water with a more reactive oxidant like H 2 O 2 or O 3 can alter the surface chemistry.The amount of oxygen incorporated into the alumina thin film during the deposition process is related to the reactivity of the oxidant and the reaction energy of the oxidant with TMA surface species.Ozone is a highly reactive oxidant, which is capable of oxidizing TMA to produce alumina with high efficiency and reduce the number of oxygen defects and surface hydroxyl groups [31,41].The O 1s and Al 2p XPS spectra will be investigated shortly hereafter to clarify the O/Al ratios and concentration of hydroxyl groups.
The C/Al ratio is also reported in table 1 and is seen to decrease when switching from H 2 O to H 2 O 2 to O 3 as the oxidant.Ozone can produce a large number of oxygen radicals, which can react with TMA molecules to produce aluminum oxide and volatile carbon-containing species.Moreover, the overall reaction energy of ozone with Al-CH 3 surface species is much greater than that of H 2 O 2 or H 2 O (shown in table 1), which strongly drives the reaction and makes it more likely to reach completion, resulting in a lower C/Al ratio [29,41].The higher C/Al observed for H 2 O-AlO x is consistent with the fact that the removal of CH 3 surface species by forming CH 4 molecules via reaction with H 2 O is less favored than ligand removal from Al ions via reaction with O 3 , due to the higher activation energy for the reaction with H 2 O [41].This results in incomplete surface reactions and a higher amount of residual carbon when using H 2 O as the oxidant [40].
The O 1s, Al .These carbon contaminants are explained by the incorporation of residual precursor that it is not completely desorbed during the growth process, as well as adventitious carbon [30].Figure 2(e) shows that the intensity of the C 1s peak decreases when replacing water with H 2 O 2 and O 3 , consistent with the lowest C/Al ratio for O 3 -AlO x reported in table 1.This decrease is due to the effective removal of trimethyl groups from TMA and desorption of by-products [42], as ozone has a higher reactivity and can result in more complete reactions compared to H 2 O 2 and H 2 O during the deposition process [32].
In order to study the uniformity, compactness and roughness, ellipsometry, SEM and AFM were conducted on the AlO x films deposited on a silicon substrate using 85 substrate oscillations.Figures 3(a The film thicknesses were obtained by ellipsometry.By plotting the film thickness as a function of the number of AP-SALD oscillations (figure 4(a)), the growth per cycle (GPC) was obtained for each oxidant.Linear trends are observed for all deposited AlO x thin films, indicating constant GPCs.All the deposited films showed higher GPCs compared to the conventional ALD process, confirming that AP-SCVD occurred due to some precursor mixing in the gas phase, i.e. the gas flow rates and reactor-substrate spacing were such that the TMA and oxidant were not perfectly isolated by the inert N 2 gas.A growth of 0.70 nm/oscillation is determined for H 2 O-AlO x , which is equivalent to 0.35 nm/cycle (one substrate oscillation corresponds to 2 AP-SALD cycles).This indicates that approximately 2-3 monolayers were deposited each cycle.The use of different oxidants resulted in different growth rates.H 2 O-AlO x was found to have the highest GPC, followed by O 3 -AlO x and H 2 O 2 -AlO x which is in a good agreement with previous studies on spatial ALD of AlO x [29].The higher growth rate for O 3 -AlO x than H 2 O 2 -AlO x is attributed to the higher reactivity of the ozone.The overall reaction energy can affect the growth rate of thin films deposited with SALD [32].It was previously reported that the overall reaction energy of two O 3 with one Al-CH 3 surface species is much greater than that observed with H 2 O 2 [41].Combined with the smallest activation energy, molecular ozone is expected to exhibit high reactivity toward oxidation of Al-CH 3 surface species.Therefore, the growth rate of the AlO x thin films at identical deposition conditions may parallel the reactivity of each oxidant, resulting in a higher GPC for AlO x deposition using O 3 [41].However, while the reactivity of H 2 O with Al-CH 3 is less than that of O 3 and H 2 O 2 [41], H 2 O-AlO x shows the largest GPC in figure 4(a).This indicates that while reaction energy  and reactivity are important, other factors such as vapor pressure and surface density of nucleation sites can also play a significant role in determining the growth rate of the deposited thin film [26].The higher GPC observed for H 2 O-AlO x could be due to multilayer water condensation on the substrate surface by physisorption [30].At low deposition temperatures (below 300 °C), water molecules tend to form multilayers instead of a monolayer and insufficient purging can lead to an additional CVD-like film growth, resulting in a higher than expected growth rate [30].The lower growth rate of H 2 O 2 -AlO x may also be attributable to the lower vapor pressure of The average RMS roughness value was evaluated from 2 to 4 separate AFM images of each film, as shown in figure 4(c), since roughness is a factor in determining the barrier properties of thin film encapsulation.A low surface roughness (RMS roughness values between 1.02 and 0.76 nm) was achieved for all AlO x films, indicating a good uniformity and smoothness over the whole scanning area.The O 3 -AlO x film is the smoothest film with the lowest RMS value of 0.76 nm, followed by H 2 O 2 -AlO x (0.93 nm), whereas the RMS roughness value for H 2 O-AlO x is 1.02 nm.The increase in surface roughness when using water as the oxygen precursor may be related to multilayered water absorption.The use of water can result in incomplete reactions and impurities on the surface, which can accumulate if not effectively removed by a sufficient nitrogen flow [32].In contrast, ozone can provide smoother thin films due to its highly reactive nature and ability to provide more complete surface coverage [50].permeation along percolation paths formed by OH− defect clusters [24].These percolation paths are formed progressively over time due to H 2 O corrosion of the film.As can be seen in figure 5, the O 3 -AlO x thin film encapsulation showed a similar oxidation pattern but at later times, indicating that the permeation rate for water molecules through the O  4(b)).A film with a higher refractive index may have more tightly packed atoms (higher density), which can reduce the number of pathways available for watervapor diffusion, resulting in a lower WVTR [30].Clearly, the rate of Ca oxidation for the H 2 O-AlO x film was considerably higher than those of the other films.The Ca films encapsulated with H 2 O-AlO x reached almost complete oxidation at 150 min, while the oxidation area of the Ca films encapsulated with H 2 O 2 -AlO x and O 3 -AlO x films changed by only 49% and 41% after 6 h, respectively.The WVTR values were obtained by analyzing the oxidized surface area of the Ca films [52].The calculated WVTR values were 5.8 indicating that all the films are good barriers to water-vapor permeation.As expected, the WVTR values decreased by using more reactive oxidants.Even though the process in this work was performed at a temperature of 130 °C, which is a relatively low temperature, by carefully controlling the deposition conditions, it is possible to achieve a high-quality TFE and minimize the WVTR.
Having established a relationship between the WVTR and oxidant, which resulted in different quantitates of surface OH− groups, it is now possible to re-consider the role of the OH− groups in the WVTR improvements reported previously when nitrogen [23] and zinc [25] were incorporated into AlO x films deposited using the same AP-SALD system (water was used as the oxidant in both cases).Ammonium hydroxide (NH 4 OH) solution (28% ammonia (NH 3 ) in water) was used as an oxidant to deposit nitrogen-doped AlO x thin films and the nitrogen doping was controlled by varying the NH 4 OH bubbling rate [23].Diethyl zinc [Zn(C 2 H 5 ) 2 , (DEZ)] was used as the zinc precursor to deposit zinc aluminum oxide.The flows from the DEZ and TMA bubblers were mixed and co-injected to produce a homogeneous film and the zinc concentration was controlled by varying the DEZ flow rate, while keeping the TMA bubbler flow rate fixed [25].Both nitrogen and zinc incorporation were found to reduce the concentration of OH− groups in the AlO x films, increase the film compactness (as measured by ellipsometry), and reduce the WVTR.However, it was unclear to what extent the improvements in WVTR were due to the reduction in OH− groups and associated waterpermeation pathways versus the presence of these additional atoms.In the case of nitrogen doping, it was proposed that nitrogen atoms could act as energy wells that trap permeating water molecules, and it was proposed that the smaller zinc atoms could facilitate a more compact film to prevent water permeation and provide enhanced corrosion resistance [23,25].The lowest WVTR values were observed for a film with 0.28% nitrogen (0.28N-AlO x ) and a film with a Zn/Al ratio of 0.21 (0.21Zn-AlO x ) [23,25].Table 2 reports the percentages of the O 1s XPS signal attributable to OH− groups and the corresponding WVTRs for the AlO x films studied in this work, as well as the nitrogen-doped AlO x [23]and zinc-aluminum oxide films [25] from these previous studies.Data for pristine AlO x films from the nitrogen and zinc studies are also provided in the table for reference.It is seen in table 2 that in this study when the OH− XPS signal was reduced from 29% (H 2 O-AlO x ) to 19% (H 2 O 2 -AlO x ), the WVTR decreased by approximately 71% from 5.8 × 10 −4 to 1.7 × 10 −4 g m −2 d −1 .In the nitrogen-doped AlO x , a higher fraction of OH− groups was observed in the O 1s XPS spectra, which may be attributable to the synthesis conditions at different times (e.g.relative humidity).But it is seen that the addition of 0.28 at.% nitrogen into the alumina lattice leads to a relatively small reduction in the fraction of OH− group in the O 1s XPS spectra from 41% to 36%, but the WVTR decreased from 7.8 × 10 −4 to 5.7 × 10 −4 g m −2 d −1 , a smaller transmission rate than was observed when H 2 O was used as the oxidant in this study and the OH− contribution was only 29%.This suggests that the WVTR reduction in the nitrogen-doped AlO x cannot be attributable solely to the reduction in OH− groups and that the nitrogen energy wells discussed in the previous study may be equally effective in terms of impeding H 2 O diffusion through the TFE.Similarly, it is seen in table 2 that the zinc-aluminum oxide film displayed a very modest 13% reduction in the OH− XPS signal but a dramatic 82% reduction in WVTR compared to the pristine AlO x sample in that study.This confirms that the WVTR reduction in the Zn-AlO x was not solely attributable to the reduction in surface OH− groups and that the zinc plays an active role by reducing water corrosion and increasing the refractive index.Thus, while it is seen in this study that selecting an appropriate oxidant can reduce the WVTR of AlO x films deposited using an AP-SALD system, additional dopants and alloying elements are also beneficial [23,25].However, further detailed studies are warranted to validate and confirm the extent of these contributions.However, after deposition of the O 3 -AlO x TFE, the opencircuit voltage (V OC ) decreased from 1.03 to 0.98 V in figure 6(c), while there was a minimal change in the FF and short circuit current (J SC ).This indicates that there is a possibility that the Ag electrode can be damaged by the highly reactive ozone [54].Ozone may react with the Ag electrode, leading to the oxidation of Ag.Previous studies have reported damage resulting from exposure to certain ALD precursors [8], which is expected to be more severe compared to SALD.The slow deposition rate of conventional ALD has limited its use for the deposition of TFE layers on PSCs due to the need for prolonged exposure steps.In contrast, the much shorter AP-SALD times used in this work allow higher deposition temperatures (e.g.130 °C) and more reactive precursors to be used without destroying the device stack [15].
To study the performance of the TFE layers developed in this work, p-i-n PSCs were coated with 60 nm of H 2 O-AlO x , H 2 O 2 -AlO x , and O 3 -AlO x and then the encapsulated cells and control cells without an encapsulation layer were exposed to accelerated test conditions (ISOS-D-3: dark, 65 °C, 85% RH) [36] and the PCEs of the cells were monitored over extended periods.Figure 6(d) presents the normalized PCEs as a function of time.The stability of the devices increased markedly by applying an encapsulation layer.For example, the PSC without encapsulation in figure 6(d) degraded to 80% of its original efficiency in 55 h (T 80 = 55 h) [36], due to moisture permeating into the device and causing accelerated degradation of the perovskite layer and its interface with charge transport layers [55].In contrast, all encapsulated PSCs in figure 6(d) displayed significantly longer T 80 values greater than 196 h, as shown in figure 6(e).The AlO x TFE layers are expected to slow down the ingress of moisture and oxygen from the ambient and the egress of volatile species from the device.The most effective encapsulation was offered by O 3 -AlO x as the efficiency for that device reduced by just 20% over a period of 312 h.In contrast, the efficiency for the PSC encapsulated with H 2 O-AlO x decreased by 60% in 312 h.The T 80 value for the PSC encapsulated with H 2 O 2 -AlO x is comparable to that of the PSC encapsulated with O 3 -AlO x and significantly better than that of the H 2 O-AlO x -encapsulated device.This indicates that the H 2 O 2 -AlO x and O 3 -AlO x films can more effectively block moisture diffusion and provide good protection for PSCs.

Conclusion
In this work, we used an AP-SALD system operated in AP-SCVD conditions to produce uniform AlOx thin films with different oxidants (H 2 O, H 2 O 2 , and O 3 ) for encapsulation of PSCs.At identical process conditions, H 2 O-AlO x thin films showed the highest growth rate (GPC = 0.35 nm/cycle) followed by O 3 -AlO x (GPC = 0.30 nm/cylce) and H 2 O 2 -AlO x (GPC = 0.26 nm/cycle).The high GPC observed for H 2 O-AlO x , as compared to more reactive oxidants like H 2 O 2 or O 3 , could be due to multilayer water condensation on the substrate surface at the low deposition temperature of 130 °C.Replacing water with the more reactive oxidants, H 2 O 2 and O 3 , increased the refractive index, indicating an increased film density, and reduced the RMS roughness to 0.93 and 0.76 nm, respectively, compared with 1.02 nm for the H 2 O-AlO x film.The more reactive oxidants result in more complete surface reactions that reduced the number of surface hydroxyl groups and carbon contaminants.WVTR values on the order of 10 −4 g m −2 d −1 at 65 °C and 85% RH were measured and the WVTR values decreased by using the more reactive oxidants, which was attributed to the lower concentration of OH− defects.The rapid nature of the AP-SCVD technique enabled deposition of the TFE layers without damage to the heat-sensitive materials used for PSC fabrication.When applied to p-i-n PSCs, deposition of the TFE layer using H 2 O and H 2 O 2 at 130 °C had no obvious effect on the solar cell performance.However, the O 3 -AlO x TFE deposition caused a slight decrease in the V OC , indicating that the Ag electrode may be damaged by exposure to the highly reactive ozone.The most effective encapsulation was offered by the O 3 -AlO x as the efficiency for that device reduced by just 20% over a period of 312 h at 65 °C and 85% RH.However, the T 80 value for the PSC encapsulated with H 2 O 2 -AlO x was comparable (288 h) and significantly better than the H 2 O-AlO x -encapsulated device.This indicates that both H 2 O 2 -AlO x and O 3 -AlO x provided good protection but H 2 O 2 did not cause any damage to the PSC.Overall, the H 2 O 2 -AlO x film has an ultralow WVTR value (1.7 10 −4 g m −2 d −1 in ISOS-D-3 conditions) and may be a more practical and effective choice due to its good barrier performance, compatibility with the perovskite layer, and commercial availability.Thus, the AP-SCVD AlO x thin films presented here provide a promising new PSC-encapsulation route, not only for lab-scale studies but also for highthroughput commercialization efforts.

Figure 1 .
Figure 1.Scheme of simultaneous metal precursor and oxidant injection in AP-SALD reactor, where the precursor half-reaction zones are separated by inert gas curtains.By moving the substrate horizontally underneath the reactor, two half reactions take place sequentially on the substrate.
2p and C 1s XPS spectra of the 60 nm thick AlO x deposited with different oxygen precursors are shown in figure 2. Deconvolution of the O 1s peaks by Gaussian fitting in figures 2(a)-(c) revealed peaks at 530.6 and 532.2 eV, which correspond to lattice oxygen (i.e.Al-O) and non-lattice and surface oxygen (e.g.O-H, hydroxyl groups), respectively [42].The Al 2p spectra (figure 2(d)) showed a single peak with a binding energy of 74.3 eV corresponding to the Al-O bond, consistent with the 530.6 eV peak in the O 1s spectra (figures 2(a)-(c)) [43].The fraction of the O 1s signal attributable to Al-O and O-H components was calculated for the H 2 O-AlO x , H 2 O 2 -AlO x and O 3 -AlO x thin films based on the area under the corresponding XPS peaks and is listed inside each plot.The Al-O fraction of the O 1s spectra is higher for the H 2 O 2 -AlO x film (81%) than the H 2 O-AlO x film (71%) (figures 2(a), (b)) and further increased to 87% when using O 3 as the oxygen precursor (figure 2(c)).It is known that thermal deposition at low temperatures using H 2 O as the oxidant usually results in a large amount of OH− groups on the film surface, due to incomplete reaction of the precursors and the high desorption energy of water at low temperatures [44, 45].Stronger oxidants (H 2 O 2 or O 3 ) can more efficiently oxidize the TMA precursor, leading to the formation of a higher proportion of surface Al-O groups and fewer surface O-H hydroxyl groups [30], which is consistent with the observed reduction in O/Al %.Considering now the C 1s spectra (figure 2(e)), the presence of several components was noted.The main components of the C 1s signals are aligned at 284.4 eV, which is attributed to the presence of hydrocarbons physiosorbed on the sample surface.The second peak at 286 eV can be assigned to C-O bonds.In addition, a non-negligible third component lies at a binding energy of 289 eV and can be assigned to the presence of carbonate species [46] )-(f) show SEM and AFM images of the H 2 O-AlO x , H 2 O 2 -AlO x and O 3 -AlO x films.All films are smooth, pinhole-and crack-free, indicating a high quality well-suited for TFE.Figures3(a) and (d) showed some randomly distributed large particles on the surface of H 2 O-AlO x , which could result from gradual powder formation on the underside of the AP-SALD reactor head, indicating that some mixing of the precursors in the gas phase may have occurred, which is discussed further below.

Figure 2 .
Figure 2. O 1s XPS spectra of (a) H 2 O-AlO x , (b) H 2 O 2 -AlO x , and (c) O 3 -AlO x thin films.The fraction of the O 1s XPS signal attributable to Al-O and O-H components is reported inside the plots.(d) Al 2p and (e) C 1s XPS spectra of the AlO x thin films deposited using different oxygen precursors.

Figure 4 .
Figure 4. (a) Film thickness as a function of the number of AP-SALD oscillations for different oxygen precursors on silicon substrates.Corresponding growth per cycles are listed inside the plot.(b) Refractive indexes, and (c) surface roughness for different oxygen precursors (measured for 60 nm thick films deposited on silicon substrates).
H 2 O 2 [47].Considering the Langmuir adsorption isotherm, the maximum surface coverage of chemisorbed OH− groups might strongly depend on the vapor pressure of H 2 O or H 2 O 2 [47].Since the vapor pressure of H 2 O 2 (7.33 Torr) is lower than that of H 2 O (9.2 Torr), a lower density of chemisorbed OH− groups is expected to be generated when using H 2 O 2 instead of H 2 O [47].The refractive indices at a wavelength of 632.8 nm were measured by ellipsometry to be 1.56, 1.63 and 1.67 for the H 2 O-AlO x , H 2 O 2 -AlO x and O 3 -AlO x , respectively, as shown in figure 4(b).These match well with the 1.53-1.78range reported previously for Al 2 O 3 [43].H 2 O 2 -AlO x and O 3 -AlO x thin films showed a higher refractive index than H 2 O-AlO x , indicating an increased film density that is associated with an improved film quality [48], which is in a good agreement with the lower O-H and carbon contents of the H 2 O 2 -AlO x and O 3 -AlO x films observed by XPS.In general, the refractive index of a film depends on its density, which can be increased by reducing the incorporation of CH 3 by-products and other impurities, such as surface hydroxyl groups [49].

2. 5 . 2 .
Water vapor transmission rate measurements.Highcontrast black-and-white images of the Ca films encapsulated with the AlO x thin films after different testing durations in ISOS-D-3 conditions (dark, 65 °C, 85%RH) are shown in figure 5.The defects observed in figure 5 are likely caused by water corrosion of the AlO x thin films, which produced oxidation areas that grew radially versus time.This is similar to the mechanism described in recent studies for Al 2 O 3 ALD TFE, where H 2 O transport through the film occurs by

Figure 5 .
Figure 5. High-contrast black-and-white images of 1.8 cm ´1.8 cm Ca films encapsulated with (a) H 2 O-AlO x , (b) H 2 O 2 -AlO x and (c) O 3 -AlO x TFE with a thickness of 60 nm after storage in ISOS-D-3 (dark, 65 °C, 85% RH) conditions for increasing time intervals.

Figure 6 .
Figure 6.J-V characteristics of p-i-n PSCs before and after deposition of 60 nm (a) H 2 O-AlO x , (b) H 2 O 2 -AlO x and (c) O 3 -AlO x TFE layers.(d) Normalized PCE versus time and (e) T 80 values of p-i-n PSCs encapsulated with different thin film encapsulation kept at ISOS-D-3 (dark, 65 °C, 85%RH).

Table 1 .
Deposition conditions, compositions, and overall reaction energy of AP-SCVD AlO x thin films deposited using different oxidants, as determined by x-ray photoelectron spectroscopy.
Sample AP-SALD deposition conditions Film composition Overall reaction energy (kcal mol −1 ) [41] TMA precursor bubbling rate (SCCM) TMA carrier gas flow rate (SCCM) Thin film characterization.XPS was performed to clarify the chemical composition of 60 nm thick H 2 O-AlO x , H 2 O 2 -AlO x and O 3 -AlO x thin films deposited using 85 AP-SALD oscillations in 5 min (table 1).For the samples prepared with O 3 and H 2 O 2 , we found O/Al ratios of 1.38 and 1.42, respectively, which are lower than that of the H 2 O-AlO x 3 -AlO x TFE was lower than for the H 2 O-AlO x and H 2 O 2 -AlO x TFE, indicating a lower WVTR value.This is attributed to the lower concentration of OH− defects in the O 3 -AlO x film as compared to the H 2 O 2 -AlO x and H 2 O-AlO x , consistent with the XPS results in figure 2. A lower concentration of surface hydroxyl groups reduces the probability of percolation paths forming and thus decreases the WVTR [51].Moreover, replacing water with a more reactive oxidant (O 3 or H 2 O 2 ) increased the refractive index (figure

Table 2 .
Comparison of surface hydroxyl groups and WVTRs for AlO x films from this work and previous studies.
AlO x , H 2 O 2 -AlO x , and O 3 -AlO x TFE layers are shown in figures 6(a)-(c).The power conversion efficiencies (PCEs) and fill factors (FFs) of the PSCs are listed inside figure 6(a)-(c).The J-V curves are nearly identical for PSCs encapsulated with H 2 O-AlO x and H 2 O 2 -AlO x films in figures 6(a), (b), indicating that the deposition of the TFE layer using H 2 O and H 2 O 2 at 130 °C had no obvious effect on the performance of the solar cells.The FA 0.92 MA 0.08 PbI 3 perovskite layer used in this study is expected to be unstable above approximately 150 °C [53].
doped TiO x and AlO x insulator layers Adv.Electron.Mater.7 1-15 [43] Alshehri A H et al 2019 Quantum-tunneling metal-insulatormetal diodes made by rapid atmospheric pressure chemical vapor deposition Adv.Funct.Mater.29 1805533 [44] Muñoz-Rojas D, Maindron T, Esteve A, Piallat F, Kools J C S and Decams J M 2019 Speeding up the unique assets of atomic layer deposition Mater.Today Chem.12 96-120 [45] Mione M A, Vandalon V, Kessels W M M and Roozeboom F 2022 Temperature study of atmospheric-pressure plasmaenhanced spatial ALD of Al 2 O 3 using infrared and optical emission spectroscopy J. Vac.Sci.Technol.A 40 062407 [46] Edy R, Huang X, Guo Y, Zhang J and Shi J 2013 Influence of argon plasma on the deposition of Al 2 O 3 film onto the pet surfaces by atomic layer deposition Nanoscale Res.Lett.8 1-9 [47] Lee W J et al 2019 Comparative study of the electrical characteristics of ALD-ZnO thin films using H 2 O and H 2 O 2 as the oxidants J. Am.Ceram.Soc.102 5881-9 [48] Dieter Mergel D M and Jerman M J M 2010 Density and refractive index of thin evaporated films Chinese Opt.Lett.8 67-72 [49] Lee J G, Kim H G and Kim S S 2013 Enhancement of barrier properties of aluminum oxide layer by optimization of plasma-enhanced atomic layer deposition process Thin Solid Films 534 515-9 [50] Azcatl A, Wang Q, Kim M J and Wallace R M 2017 Al 2 O 3 on WSe 2 by ozone based atomic layer deposition: nucleation and interface study APL Mater.5 1-7 [51] Lv Y et al 2018 Low-temperature atomic layer deposition of metal oxide layers for perovskite solar cells with high efficiency and stability under harsh environmental conditions ACS Appl.Mater.Interfaces 10 23928-37 [52] Choi J H et al 2007 Evaluation of gas permeation barrier properties using electrical measurements of calcium degradation Rev. Sci.Instrum.78 064701 [53] Jiang Q et al 2019 Surface passivation of perovskite film for efficient solar cells Nat.Photon.13 460-6 [54] Guerrero A et al 2015 Interfacial degradation of planar lead halide perovskite solar cells ACS Nano 10 218-24 [55] Han Y et al 2015 Degradation observations of encapsulated planar CH3NH3PbI3 perovskite solar cells at high temperatures and humidity J. Mater.Chem.A 3 8139-47