An Interlaboratory Study on the Stability of All-Printable Hole Transport Material–Free Perovskite Solar Cells

Comparison between different laboratories on long-term stability analyses of perovskite solar cells (PSCs) is still lacking in the literature. This work presents the results of an inter-laboratory study carried out between 5 laboratories from 4 countries. Carbon-based PSCs were prepared by screen printing, encapsulated and sent to different laboratories across Europe to assess their stability by the application of three ISOS aging protocols: (a) in the dark (ISOS-D), (b) under simulated sunlight (ISOS-L) and (c) outdoors (ISOS-O). Over 1000 hours stability is reported for devices in the dark, both at room temperature and at 65 °C. Under continuous illumination at open circuit, cells survived only for few hours, although they recovered after being stored in the dark. Better stability is observed for cells biased at maximum power point under illumination. Finally, devices operate in outdoors for 30 days, with minor degradation, in two different locations (Barcelona, Spain and Paola, Malta). Our findings demonstrate that open circuit conditions are too severe for stability assessment and that the diurnal variation of the PV parameters reveals performance to be strongly limited by the fill factor, in the central hours of the day, due to the high series resistance of the carbon electrode.

In the wide range of possible device architectures and material combinations demonstrated so far for PSCs, carbon-based HTM-free PSCs (C-PSC) are one of the most promising, in terms of ease of manufacture, [15] environmental impact, [16] and long-term stability. [17] Consisting of an all-printable triple mesoscopic stack, i.e. titania (TiO2) electron transport layer, zirconia (ZnO2) insulating layer and carbon-based back electrode, they show great potential as a low-cost PV technology for industrial production. Efforts have been concentrated to improve the manufacturing process [18][19][20][21][22][23] and large C-PSC modules have already been reported by different groups, with power conversion efficiency (PCE) ranging between 6 and 11%. [4,[6][7][8][9][10] When considering small cells (≤ 1 cm 2 of active area), C-PSCs lag behind other PSC architectures that have recently exceeded 25% efficiency. [3] Regardless, a certified PCE as high as 12.84% has been reported for the MAPI infiltrated TiO2/ZrO2/C stack [15] while the record PCE ranges between 16% for a triple cation perovskite absorber, infiltrated in the same triple mesoscopic structure, [26] and 17% for a PIN structure, also endowed with a triple cation perovskite plus a nickel oxide layer between the insulator and the carbon electrode. [27] An important milestone in their development was the addition of 5-AVAI (5-ammonium valeric acid iodide) to the perovskite precursors' solution: it has been proved to induce the formation of a peculiar multi-dimensional 2D/3D perovskite junction, i.e. AVA-MAPI, featuring both the enhanced stability of 2D perovskites and the broad absorption and excellent charge transport of 3D MAPI. [17,28] The additive located at grain boundaries also passivates surface defects, limiting the oxygen induced degradation. [29] Whilst lower in efficiency, C-PSC devices, endowed with AVA-MAPI perovskite, have demonstrated remarkable stability under illumination, both indoor at 1 sun, AM1.5 (> 1 year [17] ) and outdoor (30 days, in Wuhan, China; [25] 2136 hours, i.e. 89 days, near Beijing, China [30] ). In all these reports, assessment of the stability was conducted by the manufacturer, rather than through independent verification from an alternate laboratory, which is common practice in this field. Round robin stability studies on PSC are still limited, however as the technology matures, such studies will provide vital insight into testing protocols and variability in the stability from different testing approaches. [13,[31][32][33] This work assesses the stability of C-PSCs under a variety of conditions and involves measurements at 5 independent laboratories across Europe (UK, Italy, Spain, Malta) for measurements and characterisation, following the example set by the OPV community, who has promoted and widely participated in round robins and inter-laboratory studies. [34][35][36][37] C-PSCs were manufactured by screen printing and infiltrated with AVA-MAPI solution at a single manufacturing site, encapsulated and then sent to different laboratories for characterisation in accordance with three ISOS protocols: [38] in the dark (ISOS-D), under simulated sunlight (ISOS-L), and outdoors (ISOS-O). Although this work was conducted in line with previous ISOS guidelines, which were established for OPV, [38] the work aligns with the new ISOS consensus stability standards for PSCs that has been published very recently. [39] Additional material characterisation measurements were performed to better understand the behaviour of the devices.

Results and discussion
C-PSCs used in this work consisted of a compact TiO2 (c-TiO2), deposited by spray pyrolysis on an FTO-glass covered, and three printed mesoporous layers: titania (m-TiO2), zirconia (m-ZrO2) and carbon (Figure 1 a). While the former two layers have similar porosity and are hardly distinguishable, the over 10 m thick carbon layer is made of both large graphite flakes and fine carbon black particles, as shown by the cross-section SEM (Figure 1 b). The energy dispersive X-ray spectroscopy (EDX) images (Figure 1 c) reveal that TiO2 and ZrO2 layers are about 0.8 m and 1.2 m respectively and, according to the uniform distribution of lead and iodine elements, the perovskite is infiltrated throughout the stack.

Photovoltaic performance of as-prepared devices
J-V measurements on 21 devices, masked to 0.5 cm 2 active area, returned a spread distribution of PCE values both prior to (around 4.6% in average -reverse scan) and after encapsulation (around 5.4% in average -reverse scan), but no remarkable degradation, as shown in Figure   2a. A slight improvement in performance, primarily from changes in open circuit voltage (VOC) and fill factor (FF), was noted after encapsulation along with a decrease in JSC ( Figure S1). As reported above, cells with the same structure (TiO2/ZrO2/C) and more conductive carbon electrode, infiltrated with the same perovskite, can deliver up to 12% PCE on a masked area of 0.07 cm 2 (printed area = 0.5 cm 2 ). [15] Herein, the choice of carrying out all measurements on 0.5 cm 2 masked active areas (printed area = 1 cm 2 ) explains the PCE values below 10%: reducing the masked active area from 0.5 cm 2 to 0.0625 cm 2 boosted the reverse scan PCE from 5.6% (5.2% forward scan, 5.4% stabilized at maximum power point - Figure S2) up to 9.3% (7.7% forward scan), as shown in Figure 2b. As well as for similar cells infiltrated with MAPI, [40] the dependence of the performance on the masked area is due to limitations in the conductivity of the carbon layer, affecting the series resistance, thus the fill factor (FF), as clearly shown by the slope of the J-V curve around VOC. JSC values also depended on the masked area, suggesting a non-homogenously infiltrated perovskite, possibly hindered by dense carbon flakes. [41] Despite the lower performance, 0.5 cm 2 masks were used throughout the study to allow sampling a more representative portion of the devices. IPCE spectra returned values of integrated JSC consistent with those obtained by the J-V scans under the solar simulator ( Figure   S2). Encapsulated devices were shipped by air to the other partners for stability assessment and further characterization, as detailed in Table 1. Once the cells were received and before conducting stability tests, devices were measured at the characterization laboratory, by performing J-V scans in reverse and forward directions at the same scan rate as at the manufacturing laboratory, i.e. 20 mVs -1 . No remarkable degradation due to the shipping time or, in general, to the delay between fabrication by the manufacturing laboratory and testing by the characterization laboratories was observed, as shown in Figure S3 for the cells used for the ISOS-O2 test in Barcelona (Spain).

Stability analysis in the Dark (ISOS-D)
In the dark, both at room temperature (ISOS D1) and at 65 °C (ISOS D2), the encapsulated devices proved to be remarkably stable. Interestingly, at room temperature (Figure 3, top), they suffered from an initial loss in performance, primarily due to a decrease of the VOC (Figure S4), which led to a drop of around 20% in the initial PCE in 75 hours (T80), although this stabilised afterwards without any further decrease for over 1000 hours. By contrast, when subjected to elevated temperature at 65 °C ( Figure 3, bottom), the cells experienced a 20% improvement in the average performance within the first 3-4 hours; such improvement was then retained for almost 2000 hours, without seeing any further drop in performance. In this case though, the VOC decreased in the first few hours, but this was offset by an increase in the JSC (Figure S5), leading to the overall PCE improvement. AVA-MAPI perovskite for C-PSCs were (in this case) and usually are annealed at low temperature, i.e. 50 °C, thus prolonged exposure to 65 °C led most likely led to further annealing of the absorber layer with an improvement of the perovskite crystallinity and/or interfaces quality, i.e. better contact between perovskite and mesoscopic layers.
On the devices aged in the dark, Raman spectroscopy was applied to probe the degradation products within a perovskite device stack. One of the main by-products of perovskite degradation is lead iodide (PbI2), which has a strong Raman signal, when excited with 532 nm laser [42] , and such signal intensity is sensitive to the amount of PbI2 formed in the perovskite, which can then be related to the film and/or device degradation. The PbI2 peak can be detected as long as sufficient light reaches the PbI2 formed at the interfaces and/or in the bulk of the perovskite film.
In a carbon-based PSC, the absorber layer is infiltrated all the way through the mesoporous layers. Hence, the perovskite and degradation products formed within the perovskite can be detected and monitored either from the glass/m-TiO2 side or from the air/carbon side. The penetration depth of 532 nm laser source into CH3NH3PbI3 active layer is approximately 500 nm, which is much less than the total thickness of the carbon layer (i.e. >10 m). Thus, the Raman signal from the carbon side will be indicative of perovskite and PbI2 formed near the surface of the carbon layer only. On the other hand, measuring the device stack from the glass side will give information about perovskite and degradation products formed in the mesoporous TiO2 layer only (~ 700 nm thick), i.e. in the photoactive part of the stack, where charge carriers are generated.
Raman measurements were thus performed on encapsulated cells, before and after ageing for 1200 h in the dark at room temperature and ambient humidity (ISOS-D1), and on nonencapsulated devices for comparison. The Raman spectra for the fresh encapsulated and nonencapsulated samples measured from the carbon side are shown in Figure 4 (a and b). For the encapsulated sample, typical Raman spectrum of CH3NH3PbI3 is observed with two broad and weak peaks at 110 cm -1 and 250 cm -1 , which were assigned to the vibrational and torsional modes of the methyl ammonium (MA) cations, respectively. [43] For the non-encapsulated sample, a completely different spectrum is measured, which shows sharp and intense peaks at 73 cm -1 , 96 cm -1 and 106 cm -1 indicatives of the presence of a large amount of PbI2 [44] near the carbon surface, along with non-degraded perovskite, as revealed by the broad and poorly resolved peak centered at 242 cm -1 . Hence, the perovskite film near the surface of the device has already initiated degradation shortly after fabrication due to exposure to air when the device is not encapsulated, since some PbI2 is formed alongside CH3NH3PbI3.
After aging, different spectra are observed when measuring the devices from the carbon side ( Figure 4c and Figure 4d, black lines). In the case of the encapsulated sample, no PbI2 is measured near the surface but two small peaks at 110 cm -1 and 165 cm -1 were observed, which could be due to the formation of dihydrate perovskite (CH3NH3)4PbI6·2H2O, as shown in an earlier report. [42] The presence of dihydrate perovskite suggests that small amount of water could have been trapped in the stack during the fabrication process and/or the encapsulation, both carried out in air. Still, the encapsulating glass seems to work as a barrier to environmental moisture and oxygen, preventing further degradation of the perovskite to PbI2. For the nonencapsulated sample, intense PbI2 signal is measured from the carbon side, and the perovskite Raman bands are not observed anymore (the additional peak at 215 cm -1 is an overtone of 109 cm -1 peak of PbI2 as described by Warren et al. [44] ). This clearly indicates the conversion of perovskite to PbI2 in the carbon layer, near the surface, during the aging process.
By contrast, when the aged samples are measured from the mesoporous TiO2 side (Figure 4c and Figure 4d, red lines), PbI2 is not detected for either the encapsulated or the non-encapsulated samples. Instead, the band centered at 250-253 cm -1 indicates non-degraded perovskite and the peak at 144 cm -1 matches well with anatase TiO2: [43,45] the perovskite infiltrated within the mesoporous TiO2 layer was well preserved and did not undergo any major degradation, even without encapsulation. This correlates well with the performance of the devices, as summarized in Table S1: there is no degradation of the efficiency after 1200 h of ageing for both the encapsulated and non-encapsulated samples, and even a slight improvement in reverse bias.
Indeed, Raman measurements showed that although the perovskite is degraded in the carbon electrode without encapsulation, it remains unchanged in the photoactive layer, where charges are generated (Figure S6), which explains the good stability in the dark of both the encapsulated and non-encapsulated carbon-based PSC.

Stability analisis under light irradiation at 1 sun (ISOS-L)
The high stability observed under ISOS-D1 and D2 conditions was not replicated under continuous illumination at open circuit at room temperature (in accordance with ISOS-L1 tests) nor at 65°C (ISOS-L2): in both instances, the performance dramatically dropped in few hours, regardless the temperature (Figure 5). Cells used for ISOS-L1 and L2 tests were retested one month after light soaking and they did still work, confirming that, also for this cell architecture, storage in the dark for a sufficiently long time can induce performance recovery. [12,33] In this case, the recovery was only partial; still, around the 80% of the initial value was regained, due to a slight VOC rise coupled to an irreversible JSC drop of almost 50%, that, as a consequence, boosted the FF to higher values than at the beginning of the test, mitigating the loss in performance ( Figure S7 and Figure S8).
It is well known that an open circuit bias can accelerate the degradation during light soaking tests: [12] non extracted photogenerated charges accumulate and lead to high concentrations of radicals, which, in presence of oxygen and light, degrade the device. [9,29] Also, the spectrum from a sulphur plasma lamp is broader than that of white LEDs, which do The inset shows the MPP dynamic behaviour after each JV scan: the JV hysteresis directly affects the time required to reach the steady-state MPP. As demonstrated by Pockett et al [46] for similar device structure, the slow response time under illumination can be related to the formation of 2D perovskite regions which restrict ion migration, a phenomenon that is promoted by the AVA additive.
The evolution of VOC and JSC with light intensity (Pinc) can provide information about the recombination phenomena within the device. Figure 6 shows semi-logarithmic plots of the VOC versus light intensity for the C-PSC before and after light-soaking at MPP. At low light levels (< 0.01 Wcm -2 ), the aged devices show an increased slope of the VOC (from 165 to 251 mV/dec), in particular, as demonstrated by Gouda et al, [47] at these irradiation levels the increase of VOC is limited by the TiO2 and the accumulation of the photo-injected electrons, similar to dye sensitized solar cells. This is a clear indication of a severe photoinduced degradation of the mesoporous layer.
Additional information about recombination processes can be extrapolated by the power law fit of the JSC trend at different light intensities, before and after the stability test. The reduction of the power index (γ) from 0.91 to 0.88 is an evident indication of the appearance of bimolecular recombination processes induced by the light stress that degrades the charge collection capability [48] .

Stability analysis outdoors (ISOS-O)
Outdoor measurements were carried out in two different sites, i.e. Barcelona, Spain (41° 30' Nevertheless, in both locations, as shown in Figure 7, cells were stable for several weeks (between 700 and 800 hours, i.e. around 30 days). Cells were still working when data collection ended, and IPCE spectra were remeasured indoors: little variation is observed compared to the spectrum at the beginning of the test (Figure S9). These stability measurements add further evidence of the good outdoor performance of this PSC architecture to previously published reports by other groups of one-week stability in Jeddah, Saudi Arabia, [49] 30 days in Wuhan, China [25] and nearly 90 days in location near Beijing, China, with temperatures ranging between -10 and 35 °C. [30] As the tests were carried out between April and June (late springsummer at the given latitudes) in Southern Europe, temperatures were above 20 °C during the central hours of the day and above 6 °C at night, whereas irradiation levels peaked near 1000 Wm -2 were achieved around midday in sunny days. Irradiance levels dropped during cloudy days and so did the photogenerated current, but overall, the performance was not affected, as the detrimental effect Temperature as well as variable spectral composition of sunlight during the day could explain the asymmetrical PCE trend, with the highest values in the late evening. It is worth noting that a similar behaviour (higher PCE values at the beginning and end of the day) has been reported also for NiOx/MAPI/PCBM [50] and m-TiO2/mixed cation-halide perovskite/spiro [31] devices.
As for other PSC architectures, the response of these C-PSC devices in terms of stability depend on the applied ageing conditions and can be explained by the different types of degradation, i.e. reversible or permanent, that are triggered in each case. As previously reported for mixed cation-halide perovskites on both m-TiO2 and SnO2 based cells, [31,33,51] this work demonstrates that printable C-PSC cells can degrade beyond the threshold of reversible losses under continuous illumination (faster at open circuit than when tracking the maximum power point) and have their efficiency dropping quickly, even if a partial recovery is possible upon a long enough time of storage in the dark; whereas, over light/dark cycles such as in an outdoor test, the degradation can be reversible and cells recover overnight, leading to a remarkable ~30 days stable operation in two different sites. resistance of the carbon electrode that limits the fill factor. Nonetheless, devices were stable for about 30 days of outdoor operation in two different sites, confirming that, for this architecture, the natural day/night cycling, i.e. real-world conditions, is beneficial to the long-term operation and it is a more realistic approach to assessing the stability and lifetime for C-PSC devices.

Conclusion
Since light/dark cycles can be reproduced also indoors, they should be included in long-term stability studies, as agreed on in the new ISOS protocols. [39]

Experimental Section
Cells fabrication: FTO glass substrates (TEC7, XOP) were etched using a Rofin Nd:YVO4 laser (532 nm) at a speed of 150 mm s -1 , cleaned with Hellmanex solution in deionised water, washed with deionised water and rinsed in acetone and isopropanol, before being O2 plasma treated. A 50 nm-thick compact TiO2 layer was deposited via spray pyrolysis at 300°C from a solution 1:9 of titanium di-isopropoxide bis(acetylacetonate) (Sigma) in isopropanol. The triple mesoporous stack was obtained via screen printing of commercial pastes: first, the TiO2 layer (30 NRD Dyesol, diluted 1:1 by weight with terpineol), followed by sintering at 550°C for 30 minutes; then the ZrO2 layer (Solaronix), sintered at 400°C for 30 minutes; finally, the carbon layer (Gwent Electronic Materials), sintered at 400°C for 30 minutes. A solution of AVAI, MAI and PbI2 in gamma-butyrolactone (GBL) was prepared according to Jiang et al [52] infiltrated from the top carbon electrode, percolated throughout the mesoporous stack down to the TiO2, filling the pores and, within them, slowly crystallizing into perovskite during the annealing at 50 °C for 1 hour.
Silver paint was applied to the contacts and a black tape mask to the glass side, with an aperture of 0.5 cm 2 to univocally define the active area, allowing consistency for samples measured in different laboratories. All cells were encapsulated in air using a plain glass cover and a UVcurable epoxy for edge sealing (primary encapsulation). Curing was performed under a UV lamp for few seconds, having care of exposing only the epoxy around the glass cover edges.
Cells meant for outdoor testing had wires ultrasonically soldered to the contacts to make possible the addition of a UV filter and a secondary encapsulation of waterproof silicone ( Figure S10).

Indoor ISOS tests:
Prior to shipping, the cells were characterised at the manufacturer laboratory.

Outdoor ISOS tests in Barcelona (Spain): Upon receipt of cells, JV curves were measured
indoors under a metal halide solar simulator lamp (AM 1.5G) and IPCE measurements were taken from 300 to 800 nm, both before and after the addition of a UV filter. Outdoor stability measurements were conducted on the encapsulated cells with UV filters and masks, using a 2axis tracking system, and recording forward and reverse J-V curves (sweep rate 20 mVs -1 ), temperature, and irradiance every ~45 min, when solar irradiance exceeded 50 Wm -2 as measured by the pyranometer on the solar tracker. Between measurements and when irradiance was below 50 Wm -2 cells were held at open circuit. Cells were re-measured in the lab after 30 days of outdoor measurements.

Outdoor ISOS tests in Paola (Malta):
Outdoor stability measurements were conducted on the encapsulated cells without UV filters and masks, using a fixed system with orientation 35° South, and recording forward and reverse J-V curves (sweep rate 20 mVs -1 ), temperature, and irradiance every ~30 min, when solar irradiance is approximately 800 Wm -2 as measured by the pyranometer on the surface. Between measurements and when irradiance was below 800 Wm -2 cells were held at open circuit.

Further characterization:
Morphology and thickness of the mesoporous stack was studied using a JEOL-JSM-7800F field emission scanning electron microscope (5 kV acceleration voltage, a working distance of 10 mm and a magnification of 25,000x). Energy dispersive X-ray spectroscopy mapping was used to determine the element distribution using 20 kV acceleration voltage.
Raman measurements were performed with a Renishaw Invia Raman system in backscattering configuration. A 532 nm laser and 50x objective were used (Numerical Aperture: 0.50, spot size ≈ 1 μm). A laser power of 0.3 mW and acquisition time of 5 s were used. For each sample, 100 different spots were measured over the surface and averaged to increase the signal-to-noise ratio without degrading the cell by long laser exposure. The samples were analysed from both the carbon side and glass side to probe the perovskite degradation in the carbon or mesoporous TiO2 film, respectively.
Light intensity dependent tests were performed using the commercial apparatus ARKEO (Cicci Research) endowed with a LED based illuminator (spectral emission from 450 nm to 750 nm) with 2 cm diameter of optical aperture. Devices were attached over a metal substrate to increase the thermal capacitance. No thermal control was applied to the device because the temperature variation was confined between 24 and 34 °C.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.     Figure S3 and Figure S4.              The integrated JSC did not change, the IPCE spectra modified their shape, denoting likely changes at the interfaces.