Interface Effects on the Stability of Carbon‐Electrode‐Based Perovskite Solar Cells During Damp Heating

Carbon electrode‐based perovskite solar cells (C‐PSCs) are a promising innovation in solar cell technology. For their industrial development, an innovative perovskite deposition method using inkjet printing (“ink”) is compared with the more user‐friendly drop‐casting process (“drop”). Perovskite solar cells face challenges related to sensitivity to moisture and heat, prompting an investigation into their aging behavior under damp heating conditions for both deposition processes. Results reveal significant differences in aging behavior. Performance measurements demonstrate that drop‐cast solar cells have a lifetime roughly three times longer than ink‐deposited ones, despite initial similar performance. Drop‐cast solar cells exhibit an impressive lifespan of ≈3000 h. Various characterization techniques are employed to understand the role of interfaces in C‐PSC degradation. Diffraction analyses reveal structural disparities between the two cell types, while spectroscopic measurements reveal substantial degradation of charge transfer mechanisms in ink‐deposited cells, leading to pronounced radiative recombination. These findings highlight the critical influence of deposition method and active layer thickness on device stability. Optimization is imperative to achieve stable and efficient C‐PSC devices. In summary, C‐PSCs hold great promise, but their performance and longevity are intricately linked to deposition techniques and interface properties, necessitating careful engineering for practical application.


Introduction
7][8][9][10][11][12] As shown in Figure 1, the typical architecture of these cells includes a Titanium Dioxyde (TiO 2 ) bi-layer (compact + mesoporous) as the electron transporting layer (ETL), an insulating spacer layer with Zirconium Dioxyde (ZrO 2 ) backbone, and a mesoporous carbon layer acting as the top electrode well operating without any additional hole transporting layer (HTL).This is possible thanks to the efficient hole-blocking ability of TiO 2 . [13,14]Notably, these cells can be fabricated under ambient conditions, making the process more accessible without the need for a glove box.The perovskite material, the only unstable component, can indeed be incorporated at the end of the process through drop-casting or inkjet-printing techniques, ensuring its stability before encapsulation. [9,10]In addition, the final setup has a high inherent stability also reported by other research groups. [5,7,8,15,16]ne of the challenges faced by perovskite solar cells, including carbon electrode-based variants, is their sensitivity to moisture and heat. [17]Exposure to moisture can lead to the degradation of perovskite materials, causing a decrease in performance and a shorter device lifetime.Heat can accelerate the degradation process and the moisture-related mechanisms.Damp-heating, which refers to exposing the solar cells to high temperature and humidity simultaneously, represents the most detrimental normalized aging protocol for perovskite solar cells.To mitigate the effects of damp-heating, different strategies have been proposed such as encapsulation, materials engineering, interface modification, and use of passivation layers. [18]The literature is relatively abundant concerning the encapsulation: [19][20][21][22] different materials, methods, and process can be used.For C-PSCs, a glass-toglass encapsulation is preferred, using a polymer-based sealing agent, usually processed under basic vacuum lamination.The effect of Lead Iodide (PbI 2 ) passivation particularly on carbon electrode has also been described. [23,24]In contrast, few studies have been devoted to the use of alternative materials in architectures, passivation layers, and interface modifications in C-PSCs. [11]This will be the central point of the herein proposed study.
To improve the efficiency and stability of C-PSC devices, the addition of 5-ammonium valeric acid iodide (5-AVAI) to the perovskite formulation is widely considered. [16,25]][34] To address this issue, different strategies have been scouted in literature to enhance charge extraction and transfer.For instance, the addition of an inorganic hole transport material like CuSCN (copper thiocyanate) to the perovskite solution which enables the formation of a bulk heterojunction. [34,35]In this paper, we will take the opportunity to study the impact of aging on this unusual "bump" behavior.
The present study therefore represents a highly innovative investigation in this field: the effect of the perovskite deposition process was studied to understand the interface effects.The inkjet printing process (Ink) was first selected (Figure 1) as perovskite deposition technique, for the unique ability to control the thickness of the perovskite layer with a uniform deposit.The simpler and more user-friendly drop-casting process (Drop) was also utilized for its applied interest (Figure 1) and to identify the importance of the homogeneity in layer thickness.This provides the opportunity to vary and regulate the microstructure, thickness, and proportion of the perovskite layer at the interfaces.It was crucial for understanding the underlying mechanisms.The performance and characterization tests were conducted during an 85 °C/85% R.H. damp heating aging campaign, and the impact of a potential pre-treatment step (called maturation) was also considered.This comprehensive study involved numerous solar cells and employed a diverse range of characterization methods, varying in nature and scales.By combining these approaches, we successfully elucidated the role of interfaces in the degrada-tion mechanism of these carbon electrode-based perovskite solar cells.

Photovoltaic Devices Preparation
Solar cells used in this study were sourced from Solaronix SA (Aubonne, Switzerland).The cell stack configuration consisted of a front Fluorine-Doped Tin Oxide (FTO)/glass substrate (Solaronix, 7 Ohm sq −1 ), comprising a compact TiO 2 layer (20 nm) and a mesoporous TiO 2 layer (500 nm) as the electron transport layers, followed by a 1 μm mesoporous ZrO 2 insulating layer and a 15 μm mesoporous carbon back electrode.The active area of the cells was 12.5 ×12 mm 2 .To introduce the perovskite layer, a perovskite precursor solution (5-AVA) x MA 1-x PbI 3 (reference 76 802, Solaronix) in gamma-butyrolactone (GBL) solvent was used, which contained equimolar concentrations of lead iodide (PbI 2 ) and methylammonium iodide (MAI) along with 5 mol% of 5-ammoniumvaleric acid iodide (5-AVAI).The infiltration of the perovskite precursor solution into the mesoporous layers of the stack was carried out using one of two following procedures.In the drop-casting process (drop), 5.76 μL of the perovskite precursor solution was dispensed at the center of the active area using an automated micropipette, followed by annealing at 50 °C for ≈60 min.In the inkjet printing process (ink), the perovskite precursor solution was loaded into an inkjet printer cartridge, and a total volume of 3.85 μL was deposited uniformly across the stack using a drop-on-demand inkjet printer.The samples were then annealed on a hot plate at 50 °C for 10 min to achieve the desired perovskite crystal structure.

Maturation Pre-Treatment Step and Encapsulation
In accordance with literature, [9] a maturation process was carried out to enhance the performance of devices (temperature of 40 °C and 75% relative humidity (RH) for a duration of 100-150 h).However, for a subset of the solar cells in this study, this maturation step was omitted.Subsequently, all types of cells (with or without the maturation step) were encapsulated to safeguard them during long-term characterizations such as LBIC (Laser Beam-Induced Current) and EIS (Electrochemical Impedance Spectroscopy).A glass-glass encapsulation method was employed, utilizing a Surlyn thermoplastic ionomer-based edge sealant, see Figure 1.The encapsulation procedure involved the use of a vacuum laminator equipped with an integrated pump, programmable temperature settings, and a specific duration under a pressure of 1 atmosphere.

Characterizations
J(V) measurementswere recorded under illumination with a solar light simulator Newport Oriel LCS-100 providing an AM 1.5G spectrum of 1000 W m −2 light intensity and using a SP-150 potentiostat/galvanostat of Biologic (Software EC-Lab-11.26).The power calibration (1000 W m −2 ) was systematically carried out each day using a calibrated reference silicon cell with a readout meter (Oriel, Model: 91 150 V -Series Number: 801/0871).Measurements were done with an aperture size equal to 0.64 cm 2 .The J(V) measurements were performed in reverse mode from V oc to −0.2 V. To study the "bump" phenomena, various scan rates were used going from low scan rates (1 mV s −1 ) up to fast ones (1000 mV s −1 ).The fill factor FF, short-circuit current density J sc , open-circuit voltage V oc , power conversion efficiency PCE, the series resistance R s, and the shunt resistance R sh parameters were then deduced from J(V) measurements, using a homemade program on Python software.Because of the presence of this irregularity in the J(V) curves, a specific approach has been devised utilizing a custom-built program within the Python software.This approach, as previously mentioned and depicted in Figure S1 (Supporting Information), involves a treatment that excludes the irregularity, simplifying the J(V) curve to a more typical form.The J sc value retained is the one recorded at V = 0, while the maximum power point (P MPP ) is derived from this simplified curve.Based on this parameter, the values of J MPP and V MPP (rep-resenting the current density and voltage at which the solar cell delivers its maximum power) can be extracted.As a reminder: with P inc = 1000 W m −2 , the incident light power.The results obtained for this characterization were each time verified on at least 9 different cells for each type of solar cell.

UV-Visible Spectroscopy
UV-vis diffuse reflectance and absorbance spectra were measured with a Shimadzu UV-2600 spectrophotometer equipped with an integration sphere accessory.Measurements were made out between 200 and 1100 nm at the average speed fixed by equipment (0.5 nm step).The measurements in reflectance mode can be performed on the entire solar cell, without preparation.Absorbance measurements require a preliminary step of carbon paste (CP) removal.Indeed, the opacity of the thick CP layer avoid any possible transmission of light within the entire device.This was performed using adhesive scotch peeling method and could present the disadvantage of also removing a small quantity of the perovskite layer.This technique allows the detection of the different materials constituting the active layer of the solar cell like 3D PK at 780 nm, PbI 6 octahedron or PbI 2 at 500-550 nm, and monohydrated perovskite at 380 nm.In a more convenient way, diffuse reflection measurements were used mainly to confirm the presence of the PK layer and to evaluate its bandgap value.This latter can be calculated thanks to the Tauc plot method using Kubelka-Munk function applied to reflectance spectra.

Photoluminescence (PL) Spectroscopy
The PL spectra of studied solar cells were obtained using a Shimadzu RF-6000 spectrofluorometer with a 520 nm excitation wavelength using a scan rate of 6000 nm.min −1 , an excitation slit of 5 nm, an emission slit of 5 nm and a measurement range of 600-900 nm.

XRD Analyses
XRD analyses of perovskite layers were performed on opened cells (after peeling of CP) using a PANalytical X'Pert PRO MPD X-Ray diffractometer from the consortium of common technological means (CMTC) in Grenoble between 2-theta of 8°and 50°w ith Cu-Ka source.Data were then treated on Match! software, that enables to identify the different phases, lattice parameters, and different orientations composing the perovskite layer.In addition, after an extraction of the background and an integration of all contributions, it is possible to evaluate a relative crystallinity of the studied perovskite and a PbI 2 amount present in perovskite thanks to the following equations: An attribution from the literature of the various contributions present in diffractograms is proposed in Table 1.In general, perovskite crystallizes in a tetragonal form with two main orientations: (110)/(220) and (200)/(400). [36,37]Here other orientations could be noted as indicated in the table.We found it useful to compare thedifferent proportions of these orientations for the two types of perovskite deposition before and after the maturation stage.Thus, the proportions of the different orientations are obtained as follows: For orientation 1: For orientation 2: For other orientations:

Laser-Beam Induced Current (LBIC) Mapping
A homemade apparatus was developed to acquire LBIC images of our devices.It consists of a 5 mW, 532 nm green monochromatic laser diode producing a 0.7 mm diameter beam attached to a moving X-Y stage controlled by a computer.During scanning of the cell surface, the cell output current is simultaneously measured by a source measurement unit (2602 Keithley, Cleveland, Ohio).Scanning of the whole cell active area takes ≈20 min.

Photoluminescence (PL) Imaging
Luminescence imaging experiments were performed with a homemade device composed of an opaque chamber with a camera placed 30 cm above the samples and one green light-emitting diode (LED) arrays with emission spectrum at 525 nm.A 650 nm high-pass filter from Edmund Optics (York, UK) was added to allow for spectral resolution in the setup and to minimize reflection from the excitation LEDs.Full solar cells were observed in the open-circuit condition.The acquisition time was set close to 1000 ms.As the resulting signal varied depending on acquisition conditions (time and gain), not allowing for a direct quantitative comparison of the evolutions observed, a calibration reported in a previous publication allowing for image correction was employed. [38]Any additional image treatment (cropping, colorization, or thresholding) was performed with the ImageJ freeware.

Electrochemical Impedance Spectroscopy (EIS)
Impedance measurements were performed under solar light simulator Newport Oriel LCS-100 providing an AM 1.5G spectrum and using a potentiostat/galvanostat (SP-300 with impedance measurement option, Biologic), with a potential variation of 30 mV and a range of frequencies between 1 MHz and 0.5 Hz.The measurements were performed under 1 Sun of illumination and a voltage of 0.6 V was applied to solar cell.The Nyquist plots (-Im(Z) = f(Re(Z))) were then studied and fitted by an equivalent circuit thank to the EC-Lab-11.26software.SEM observations were conducted at the Consortium of Common Technological Means (CMTC) in Grenoble using a FEG ZEISS Ultra 55 microscope.Initially, the solar cells were intentionally broken, and specific sections were extracted for subsequent preparation through cross-polishing.SEM images were captured using both secondary and back-scattered detectors.To comprehensively analyze the constituent elements within the solar cell, additional EDX (Energy-Dispersive X-ray) analyses were carried out employing a 30 mm 2 SDD (Silicon Drift Detector) detector from Bruker AXS.This enabled precise identification of all materials present within the solar cell, including perovskite, TiO 2 , ZrO 2 , and Carbon.An example of identification of different layers is presented in Figure S2 (Supporting Information).

Results and Discussion
Figure 2 illustrates the variation of PV parameters for the solar cells under study during an aging campaign conducted under 85 °C with 85% R.H. stress conditions.Initially, solar cells subjected to a maturation step (m, with empty symbols) exhibit superior performance compared to freshly processed counterparts (w0, with full symbols), see Figure 2a,d.However, there is a noticeable decline in performance during the first 220 h of aging for matured devices, merging their power conversion efficiency (PCE) with non-matured solar cells.Conversely, the non-matured cells demonstrate relative stability during this initial phase referred to as 'Phase 1′ in Figure 2a (and in Figure 2d with a logarithmic scale for better visualization).The decrease in PCE for matured cells is primarily attributed to reductions in both shortcircuit current (J sc ) and open-circuit voltage (V oc ) (Figure 2c,e, respectively), indicating the influence of the quality of the active perovskite layer and interface interactions on device performance.These findings collectively suggest that the maturation step has limited practical significance.Although there is a momentary improvement in performance, it does not substantially alter the overall energy production with aging time.
In contrast, a comparison of the processing conditions highlights a tradeoff between the Ink_ and the Drop_solar cells, obtained, respectively, using a perovskite infiltration though inkjetprinting and drop-casting.Throughout the Phase 1 aging period, the open-circuit voltage (V oc ) of Ink_solar cells consistently outperforms that of Drop_solar cells, while the short-circuit current (J sc ) remains similar for all tested cells.Consequently, there is a decline specifically in the fill factor (FF) for Drop_solar cells (depicted in Figure 2b) within the initial 220 h, whereas this parameter remains stable for Ink_solar cells, suggesting the impact of interface quality and charge transport on FF.During this phase, the series resistance (R s ) value is relatively low for all tested cells (Figure 2f).
During the Phase 2, identified in Figure 2a, a continuous decline in performance appears in Ink_solar cells regardless of their initial treatment.After 1000 h of aging, these cells exhibit insufficient performance to be considered functional.Notably, the degradation of performance is more pronounced for the matured cells.This decline can be attributed partly to a decrease in J sc , which is more significant in the matured cells, and eventually, at the end of the aging process, by a drastic increase in Rs.Conversely, the FF and V oc parameters of Ink_solar cells remain relatively stable during the entire 1000 h aging period, further highlighting the influence of interfaces on device longevity and performance.
With the Drop_solar cells, three additional aging phases were revealed after Phase 1 and labeled chronologically Phases 2, 3, and 4 (see Figure 2a).After the initial decrease corresponding to Phase 1, a surprising performance recovery was evidenced up to ≈1130 h of aging.The cells nearly returned to the initial performances (Phase 2).This is followed by a decline in performance until 2750 h of aging (Phase 3), and then a second, yet lower, improvement until 3000 h (Phase 4).The matured cells exhibit higher performances during Phase 2, while both types of cells show relatively similar behavior during Phase 3, and a slight performance increase is observed during Phase 4 for both "m" and "w 0 " Drop-solar cells.These improvements are primarily correlated with the changes in Jsc and Voc, emphasizing the interplay between interface quality, charge transport, and overall device performance over extended aging periods.The increase in performance observed in Phase 2 resulted from the enhancements in the fill factor (FF), short-circuit current (J sc ), and open-circuit voltage (V oc ) (Figure 2b,c,e).The performance decrease observed in Phase 3 is primarily due to a decline in J sc (Figure 2c).The improvement in Phase 4 is primarily driven by increases in J sc and V oc .Additionally, for Drop_solar cells, the series resistance (R s ) increases until 500 h, correlating with the decrease in power conversion efficiency (PCE).Subsequently, R s remains relatively low until 2000 h, corresponding to the reincrease in PCE.It then increases again at the end of Phase 3, explaining the PCE decline, followed by a slight decrease in coordination with the PCE increase in Phase 4 (Figure 2f).
In summary, the processing methods strongly alters the experimental results.Although both types of cells (Drop and Ink) shared the same architecture, the deposition process of the perovskite significantly influences on the long-term stability of the devices.Ink_solar cells exhibit relatively poor stability, with no functional cells remaining at 1000 h of aging.In contrast, Drop_solar cells demonstrate yields of ≈8% at 1000 h and 4% at 3000 h of aging.This demonstrates that the degradation mechanisms differ between the two types of cells.The decline in performance for inkjet-printing solar cells is mainly attributed to a decrease in J sc , which follows a more traditional degradation pattern.Conversely, for drop-casting cells, the initial performance decrease involves reductions in J sc , V oc , and FF.The subsequent recovery of performance is partly influenced by variations in V oc and FF, while the subsequent major decline is primarily caused by variations in J sc .Finally, the slight improvement observed toward the end of the aging process can be attributed to increases in both V oc and J sc .
One significant challenge associated with the configuration of carbon electrode-based devices is the requirement of high temperatures to achieve an optimal carbon layer, which necessitates a post-deposition of perovskite.While this post-deposition of perovskite offers advantages for the development of a streamlined industrial process, it introduces the need for an insulating mesoporous zirconia buffer layer between the electron transport and carbon layers.Based on previous findings, it is reasonable to attribute the difference in stability between Ink and Drop devices to the quantity of precursor solution utilized.The use of a smaller amount of precursor solution leads to a reduced reservoir of perovskite materials, making it challenging for the ZrO 2 mesoporous layer to be adequately filled over time, ultimately resulting in the formation of insulating devices.
Furthermore, there is another potential issue that arises from the relatively long distance (≈1 μm) that the holes need to travel to reach the conductive carbon cathode.This issue becomes apparent when examining the J(V) curves.These curves often exhibit noticeable hysteresis [32,39] (refer to Figure S3 Supporting Information), and their characteristics can vary considerably depending on the scan speed employed.Moreover, a distinct "bump" phenomenon may be observed to varying degrees. [9]From the literature, [9] it seems that free electrons can be extracted quickly by simple injection into the compact and mesoporous titanium oxide thin layers, while the generated holes need to proceed through the perovskite lattice within the thick mesoporous zirconium oxide before to reach the carbon contact.This is a specificity of the architecture of studied solar cells and this can lead to an imbalance in charge extraction, which often results in hysteresis behavior as well as the appearance of a "bump" in the reverse J(V) scan.To gain a better understanding of the initial aging phase, a comprehensive investigation into this bump phenomenon was conducted which is depicted in Figure 3.
When the scan speed of the J(V) measurement is adjusted within the range of 1-1000 mV s −1 , a significant variation is observed in the resulting curves.Specifically, at these extreme speeds, there is a notable increase in the short-circuit current (J sc ), such as from ≈12 to 20 mA.cm −2 for a non-matured Drop_solar cell in its initial state.No distinct bump is observed for the boundary speeds (i.e., 1 and 1000 mV s −1 ), leading to the characterization of the J(V) curves obtained at low and high speeds as representing two different states of the C-PSC devices that we will respectively call "stable" and "metastable" states.However, when intermediate scan speeds are employed, a "bump" phenomenon emerges, as depicted in Figure 3b-e).
Figure 3b-e) also reveals that the "bump" phenomenon, particularly prominent and noticeable at the initial state, completely disappears after 120 h of aging, regardless of the cell type being studied.Additionally, it can be observed that the difference between the J sc values obtained in the "stable" and "metastable" states is initially significant but becomes negligible over the course of aging.These two parameters, J sc-max and J sc-min , were thus selected and plotted against aging time for all the tested solar cells, as shown in Figure 3a).
In the initial state, the value of J sc-max are identical for all types of cells, while the J sc-min value varies.The difference between these two parameters is generally more pronounced for nonmatured cells compared to matured cells.As mentioned in a previous article, [40] one of the mechanisms observed during the maturation phase is the improvement of electron and hole transfers to the cell's electrodes.Consistent with the J(V) measurements discussed previously, phase 1 extends up to 120-200 h.The analysis of the J(V) measurements, particularly the difference between J sc-max and J sc-min , reveals that this phase can be further divided into three sub-phases (identified by vertical black lines in Figure 3a).
During the first sub-phase (up to 48 h), J sc-max remains relatively stable, hovering ≈20 mA.cm −2 , except for Ink_solar cells, which exhibit a slight decrease in this parameter.At this stage, J sc-min generally remains stable for matured solar cells, while nonmatured cells experience an increase in J sc-min, narrowing the gap between these two parameters.This phenomenon also explains the increase in fill factor (FF) and the decrease in series resistance (R s ) observed for non-matured solar cells in Figures 2c,f), respectively.This step resembles a maturation or post-maturation phase, aiming to enhance charge transfer within the cell by balancing the rates of electron and hole transfers.This explains the convergence of performance between non-matured cells and matured cells during phase 1, as demonstrated in Figure 2.
Between 48 and 120 h, a significant decrease in both J sc-max and, to a lesser extent, J sc-min is observed, resulting in all studied cells reaching identical values of J sc-min and J sc-max at 120 h.The bump phenomenon completely disappears, indicating a balanced charge transfer within the cell.This supports the previous J(V) results, indicating that matured and non-matured cells exhibit identical performance at 120 h.Therefore, this phase can be considered as the culmination of the maturation phenomenon, induced for matured solar cells through the post-treatment conducted immediately after the device fabrication process, and for non-matured solar cells through aging at 85 °C/85%RH from 100 h onward.Thus, the solar cells examined in this study, with their unique architecture, exhibit a transient phase following the manufacturing process, during which the load transfers within the cell need to be balanced.There is a slight discrepancy between Drop and Ink solar cells, which can be attributed to the fact that performance is influenced by factors beyond just J sc .
Then, after 120 h, similar values of J sc-min and J sc-max are still observed, confirming the absence of the bump phenomenon.For Drop_solar cells, the values of J sc-min and J sc-max are maintained, while a significant decrease in these parameters is observed for Ink_solar cells.These findings align well with the previous results, although there is a noticeable deviation in performance during the initial 100 h, characterized by a temporary increase in medium-speed issues.
To elucidate the degradation mechanisms then taking place during phase 2, a comprehensive set of characterization methods was employed.Alongside the conventional J(V) measurements conducted throughout the aging tests, imaging techniques were implemented on our devices to provide a more accurate assessment of the local performance of each cell.
The light beam-induced current (LBIC) measurement allow to determine the local short-circuit current induced and provides valuable insights into the spatial distribution of the performance of a solar cell (Figure 4).As expected, Ink_solar cells exhibit more homogeneous performance compared to Drop_cells at initial state.Indeed, a central pattern related to the central deposit of the "drop" can be frequently found in Drop_solar cells (at T0 or during aging).On another note, the maturation stage leads to a convergence of performance between drop and inkjet processes.During the transient phase, lasting up to 120 h, a clear improvement in average LBIC-I sc is observed, in accordance with the rise in J sc-min depicted in Figure 3a.This increase brings the LBIC-I sc values to the same level, regardless of the deposition and pre-maturation processes.Subsequently, phases 2, 3, and 4 mirror the sequences observed in J(V) measurements and indicating a strong correlation in results for both Ink and Drop devices.However, the LBIC-I sc is not as significantly affected as J(V)-J sc which can be due to the significantly lower illumination level using LBIC laser versus the solar simulator.This could minimize the impact of defaults created during aging.
For Drop_solar cells, a stabilization or slight decrease in average short-circuit current is respectively, observed for nonmatured and matured cells up to 500 h, supporting the earlier observations of PCE and J sc .An increase in this parameter is noted  until 740 h, followed by a decrease up to 2750 h of aging, with a slight increase toward the end of the aging process (3000 h), aligning with the observed variations in PCE and J sc from the J(V) measurements.The maps of the two Drop_solar cells exhibit remarkable similarity at 240 and 1000 h, confirming the excellent stability of these devices.Subsequently, a consistent decline in local current measurements is observed, with a minor stabilization toward the end of aging (between 2500 and 3000 h).These observations closely align with the previously examined J(V) measurements.
For both types of Ink_solar cells, as we transition to phase 2 after 120 h, a consistent decline in the average LBIC-I sc is observed, aligning with the findings from J(V) measurements.In addition, similar to the trends observed in macroscopic PCE and J sc , the decrease in local LBIC-I sc is much smaller for non-matured compared to matured Ink_solar cells.
Figure 5 presents the results obtained from photoluminescence (PL) imaging.With the help of a calibration step previously described in details, [38] all samples could be quantitatively compared even if measured under different conditions or after various aging times.The cells overall revealed a rather low spatial distribution at first glance.During phase 1, the PL emission remained essentially constant for all solar cells, and with the lowest intensity.Phase 2 (after 120 h) disclosed a strong difference in behavior depending on the processing method.
For Drop_solar cells, there is a stabilization in the PL signal during phase 2, associating with the trends observed in PCE.A slight decrease in PL signal at 1100 h explains the subsequent increase in PCE toward the end of this phase.Interestingly, there is no noticeable difference between non-matured and matured cells in terms of PL signal, contrary to the previous findings where matured cells exhibited better performance.It is possible that this technique may not be sensitive enough, and further spectroscopy analyses could provide more insights.It is worth noting that this technique is primarily used to assess the homogeneity of PL emissions.In terms of homogeneity, Drop_solar cells demonstrate a relatively uniform signal with only a few defects during phase 2, indicating good encapsulation quality until 2000 h of aging.However, PL mapping reveals more pronounced defects and heterogeneous PL increase, thereafter, highlighting the degradation effect and its correlation with the decrease in PCE at this stage.This is supported by the plot in Figure 5b, where a slight increase in the uncertainty bar is observed with aging time.
For Ink_solar cells, a different trend is observed.The PL imaging shows a non-homogeneous signal from the beginning of phase 2, with a large number of observable defects for non-matured cells and a significant and continuous increase in PL for matured cells up to 1000 h.It appears that these cells are more susceptible to aging.The uncertainty bars plotted in Figure 5b also confirm the increasingly heterogeneous degradation with aging time.Once again, the lower amount of perovskite used to fill the mesoporous scaffold may contribute to the formation of voids within the perovskite layer at a faster rate.Additionally, a square pattern is observed for matured solar cells, indicating preferential degradation in the illuminated area during J(V) measurements (using a 0.64 cm 2 square mask to avoid measuring potentially non-filled border areas in the case of drop-casting process).Overall, the observed variations are consistent with the previously presented results.It is important to note a difference in behavior between matured and non-matured cells, with a faster increase of PL emission all over the device for matured Ink_devices.This is in accordance with the faster decrease in performance also observed with both J(V) and LBIC characterizations.In conclusion, further investigation using the photoluminescence spectroscopy tool would be interesting to gain deeper insights into these PL results.
The results of photoluminescence spectroscopy are presented in Figure 6.Similar to other characterization techniques, the transient phases are clearly visible (Figure 6c).The differences in behavior between Ink_ and Drop_solar cells are distinctly confirmed.For Drop_solar cells, there is an initial decrease in photoluminescence (PL) intensity during phase 1 for non-matured cells, eventually reaching the value of matured cells.This confirms the occurrence of post-maturation for non-matured solar cells within the first 100 h of aging.Subsequently, a slight increase in the maximum PL signal intensity is observed throughout the aging process.In contrast, inkjet solar cells exhibit a more pronounced increase in degradation, particularly for matured cells.This indicates that the recombination mechanism is clearly affected by aging in this type of cells.Additionally, the good stability of Drop_solar cells is also supported by this technique.
Regarding the variation of the emission peak's wavelength to the maximum intensity, a significant redshift is observed during the initial aging period of the first 500 h for all solar cells.This redshift likely indicates an improvement in the crystalline microstructure of the perovskite during phase. [41]o establish a connection between these results and the potential degradation of the active layer, UV-vis spectroscopy analyses were conducted, Figures 7 and 8. Two types of analyses were carried out: reflectance analyses, which allowed for continuous monitoring without damaging the sample, and absorbance analyses, which required opening the cell and removing the carbon layer.Although the latter analyses were more informative, they could not be performed as frequently during the aging campaign due to their destructive nature.
In Figure 7, subplots a) and b), the reflectance spectra clearly show the bandgap of the 3D MAPbI 3 perovskite, which predominantly forms the active layer, at ≈780 nm.The Tauc plots shown in the inset reveal that the bandgap energy remains constant at 1.59 eV throughout the aging process, irrespective of the solar cell type.This may appear somewhat surprising, as the degradation of perovskite should lead to a change in the bandgap. [42]However, the substantial difference in reflectance around the MAPbI 3 transition indicates that MAPbI 3 is consistently present in significant quantities at the end of all aging campaigns.
By employing this technique, we were able to observe a notable increase in reflectance across the entire spectral range (200-1100 nm) during aging.Since the reflectance enhancement is observed throughout the entire range, it does not indicate the presence of degradation products, but rather a modification in the device's filling.It is worth noting that a Glass/TiO 2 /ZrO 2 /C stack without perovskite exhibited higher reflectance levels compared to both (drop) and (ink)_filled stacks at T0 (refer to Figure S4, Supporting Information).
To compare the different studied cells, the reflectance level at 600 nm was thus plotted against aging time for all devices (Figure 7c).Initially, a similar slope was observed for all cells until 500 h.After that, this parameter stabilized for the Drop_solar cells, coinciding with a re-increase in power conversion efficiency (PCE).In contrast, the Ink_solar cells continued to exhibit an upward trend.Additionally, the increase in reflectance was more pronounced for aged Ink_solar cells, confirming their higher susceptibility to aging.These findings align with previous results.
The higher reflectance level observed in Ink_devices compared to Drop ones at the start (T0) can be attributed to the lower perovskite filling in inkjet devices.Subsequently, the degradation or amorphization of perovskite in Ink_solar cells leads to a continued increase in reflectance, eventually converging with the reflectance level of a stack without perovskite.The low stability of Ink_solar cells is further supported by the drop in performance observed during the aging campaign, accompanied by a significant decrease in short-circuit current density (J sc ), which can be partially explained by a substantial reduction in the perovskite layer.
Notably, at 1000 h, reflectance spectra of Ink devices exhibit the emergence of perovskite degradation by-products, indicated by a transition with a 400 nm bandgap corresponding to hydrated MAPbI 3 . [43,44]To observe the PbI 2 transition, the aging campaign should continue well beyond the point of reaching null PCE values, as demonstrated in Figure S5 (Supporting Informa-tion), which shows the reflectance spectra at 1600 h for Ink_solar cells.
Drop_solar cells demonstrate greater stability compared to Ink_solar cells, despite having a similar device architecture.The discrepancy can be attributed to the different deposition processes of perovskite.In Ink_solar cells, the perovskite layer thickness is significantly lower, and the resulting microstructure appears to be more defective compared to Drop_solar cells.Consequently, it can be inferred that this thinner perovskite layer is more susceptible to aging.
To gain further insights into the performance variations, additional UV-vis analyses were conducted on opened solar cells after removing the carbon layer.Figure 8 illustrates the absorbance spectra obtained at different aging times for both Drop_ and Ink_solar cells.The main transition at 780 nm concerns the 3D MAPbI 3 perovskite phase and there is no variation observed for this transition, corresponding to the dominant phase in the active layer, until the end of the aging period.Nevertheless, a more deepened analysis of this spectra enables to identify other products present in the active layer of studied solar cells.For Drop_solar cells at T0, matured devices exhibit the presence of monohydrated perovskites (at ≈350-400 nm), contrary to nonmatured cells.In the case of Ink_solar cells at T0, the formation of such phases is not observed, except for an intensified peak of PbI 2 around 550 nm in matured cells.These identifications are based on several literature sources. [10,27,41,43,44]After 500 h of aging, the perovskite spectra in Drop_solar cells flatten out, approaching the profiles observed in Ink_solar cells at T0.This trend is consistent with a less compact perovskite phase upon aging, as supported by the earlier increase in reflectance levels between 0 and 500 h of aging.In addition, the monohydrated perovskites vanish in matured Drop_solar cells, while PbI 2 remains present and slightly increases in concentration during aging across all cell types (zoomed-in section presented in Figure 8a shows after 500 h of aging an increase of absorbance in the 500-550 nm range for a normalized perovskite transition at 780 nm).In the case of Ink_solar cells, a significant increase in PbI 2 oc-curs between 0 and 500 h of aging, followed by the presence of monohydrated MAPbI 3 before 400 nm at the end of aging.
Figure S6 (Supporting Information) showcases the XRD analysis results conducted on Drop and Ink solar cells throughout the aging campaign.By carefully examining the diffractogram, it is possible to assess the relative crystallinity rate and the proportions of specific phases such as perovskite monohydrate and PbI 2 , as shown in Figure 9.
For Drop_solar cells, the crystallinity rate initially decreases up to 500 h, then gradually increases to a value similar to the initial state at 1000 h, followed by a decline up to 3000 h.The impact of maturation is minimal.Additionally, there is an observable increase in the proportion of the monohydrated perovskite phase at 500 h, followed by its complete disappearance from 1000 h until the end of the aging process.Therefore, the decline in performance observed for Drop_solar cells from 120 to 500 h can be attributed to a reduction in crystallinity and the appearance of the monohydrated perovskite phase.Subsequently, a phenomenon Figure 8. UV-vis absorbance spectra performed on opened cells after carbon layer peeling as a function of aging time for a) (drop) solar cells and b) (ink) solar cells (To obtain these spectra, some solar cells were opened, and the carbon electrode peeled off at best.To limit the contribution of carbon removal step to the spectra, and as the perovskite bandgap appears to change little during aging, the absorbance values of each spectrum were normalized according to the perovskite jump observed at 750 nm.In this way, it is easier to observe other degradation products.). of perovskite microstructure reorganization enables an increase in crystallinity, resulting in improved performance between 500 and 1000 h, reaching a level comparable to the initial performance.According to literature, the formation of monohydrated perovskite may be reversible. [43]Hence, it is plausible to hypothesize that the aging temperature of 85 °C facilitates the recrystallization of MAPbI 3 perovskite from the hydrate, potentially due to water vapor captured by the Surlyn gasket.After 1000 h, the formation of PbI 2 continues to increase, reaching ≈8% and 12% for non-matured and matured cells Drop_devices, respectively, after 3000 h of aging.This corresponds to the decreasing crystallinity of perovskite.The final drastic decline in power conversion efficiency (PCE) can be thus attributed to the significant conversion of perovskite into the inactive PbI 2 compound.
A distinct XRD behavior is observed for Ink_solar cells compared to Drop_solar cells.Notably, there is a continuous decrease in the amount of crystallinity, which can account for the persistent decline in performance and J sc after phase 1.When comparing the inkjet-printing and drop-casting processes, the crystalline ratio of perovskite in both cell types is equivalent at 1000 and 3000 h of aging, respectively, for Ink_ and Drop_solar cells.Surprisingly, despite similar crystalline rates, Drop-solar cells demonstrate a considerably longer lifespan compared to Inksolar cells.Furthermore, no crystalline monohydrated perovskite formation is observed in Ink_solar cells, and there is minimal formation of the degradation product PbI 2 in its crystalline form.Although these compounds are detected in UV-vis spectroscopy during aging, they appear to be almost absent in the crystalline state and should be more amorphous.
As discussed earlier, the mesoporous structure of TiO 2 allows perovskite to crystallize in various orientations (see Figure 9c,d).For both types of cells, whether in a non-matured or matured state, perovskite predominantly exhibits orientation 1 at T0.At 500 h of aging, there is an increase in this contribution for the four different types of studied devices.Consequently, the decrease in the crystallinity rate observed between 0 and 500 h is primarily due to the disappearance of orientation 2 and other orientations.At 1000 h of aging for both Drop_ and Ink_solar cells, there is however this time a subsequent decrease in the percentage of orientation 1 more pronounced for Drop_devices, with orientation 2 and other orientations becoming more prominent.These changes in orientation proportions correlate with the divergent alterations in the crystallinity rate between 500 and 1000 h for the two types of cells studied.It suggests that the previously hypothesized recrystallization of MAPbI 3 perovskite from the hydrate in Drop_solar cells predominantly leads to orientations other than orientation 1.In the same aging time range, for Ink_solar cells, a reduction in relative crystallinity is observed along with the appearance of monohydrated perovskite and PbI 2 in a relatively high proportion but mainly in an amorphous form, which primarily affects orientation 1.At 1000 h of aging, Ink_devices are close to their end of life with a perovskite crystallinity rate of ≈10%, while Drop_devices possesses performances close to their initial value with a perovskite crystallinity rate of ≈20%.After 3000 h of aging for Drop_solar cells, an increase in the proportion of orientation 1 is again observed, indicating that the newly recrystallized perovskite from the hydrate is the first to degrade into PbI 2 .Between 1000 and 3000 h, a significant decrease in the crystallinity and a notable increase in the proportion of crystalline PbI 2 are observed.However, the proportions of perovskite orientations have returned relatively similar to that of the initial state.
To gain a deeper understanding of the changes in solar cells during damp-heating stress conditions, some impedance spectroscopy measurements were also conducted.This widely used technique consists in studying the complex impedance of the investigated systems using Nyquist diagrams.To analyze the experimental data, various equivalent circuits were simulated, which helped to establish connections between different mechanisms related to ionic and electronic conduction phenomena occurring within the active layer and its interfaces.The solar cells were examined under non-open circuit (NOC) conditions, that is, under illumination (1 SUN) with a fixed potential of 0.6 V.These conditions were optimized to be non-destructive for the solar cells studied.Three cells of each type were closely monitored throughout the aging process.Figure S7 (Supporting Information) illustrates the raw results from one cell, while Figure 10 displays the simulation results, which present averages from three cells.The Nyquist graphs depicted in Figure S7 (Supporting Information) reveal two semi-circles with distinct characteristic times, corresponding to low and high frequencies.
Among the available options in the literature, a specific equivalent circuit was chosen for data adjustment.[47][48] The first parameter, R1, represents the ohmic contributions of contacts and wires.The equivalent circuit includes two capacitances, C1 and C3, and two resistances, R2 and R3.C1 is linked to the rapid relaxation of charges and dominates at high frequencies (> 1 KHz), signifying the bulk response of the perovskite layer.On the other hand, C3 is associated with slower charges movement and is evident at lower frequencies, indicating interfacial polarization due to charge accumulation at the surfaces surrounding the perovskite.The dielectric constant of the perovskite layer, , is utilized to estimate the geometrical capacitance of the parallel plate capacitor, given by C1 =  0 /d, where d is the layer thickness, and  0 is the permittivity of free space.C3 is associated with the accumulation of space charges at the interfaces.R2 and R3 represent the low and high-frequency resistances connected to the recombination process within the active layer and at the perovskite's interface.The total recombination resistance, R rec , can be determined by adding R2 and R3.Under NOC conditions, an exponential dependence of all parameters (resistances and capacitances) is observed at high DC potential but becomes flat as the experiment approaches SC conditions, suggesting that the device is ruled by the shunt resistance at fixed light intensity and low applied bias.The transition between the two regimes occurs at voltages right below the maximum power point (MPP) of the corresponding J(V) curves.At high voltage, the behavior is similar to that in OC conditions.Thus, to achieve conditions as close as possible to operating conditions, an intermediate potential value (0.6 V) has been chosen.This is below the V oc value of ≈0.8-0.9V.
As anticipated, the resistance R1, which represents contact and wire resistances, is quite insignificant, measuring ≈20 Ohm cm 2 , and remains relatively consistent across the various cell types investigated.The only distinction lies in the initial state, where unmatured Ink solar cells exhibit a higher value that subsequently decreases which is likely a result of maturation occurring in the first stage of aging as previously demonstrated.Throughout the aging process, this parameter remains then relatively stable, except for the Drop_matured solar cells, which display a notable increase toward the end of the study, indicating potential contact degradation.
Regarding the recombination resistances within the perovskite layer (R2) and at its interfaces (R3), it is noteworthy that the R2 values remain stable during the aging process for Drop_solar cells, indicating limited degradation of the perovskite layer.On the other hand, there is a notable and significant increase in R3 as the cells age.However, for Ink_solar cells, the variations in recombination resistances are much more pronounced.They exhibit a very significant increase in R2 toward the end of the aging period, along with significantly higher R3 values compared to those observed in Drop_solar cells.To fully interpret these findings, it is essential to establish correlations with other impedance spectroscopy measurements and complementary characterization techniques.Examining the data in conjunction with these additional analyses will provide a more comprehensive understanding of the changes occurring in the solar cells during aging.
The C1 value remains remarkably consistent across all studied solar cells, being close to 0.08 μF.cm −2 in the initial state.Interestingly, despite the possibility of a thinner active layer in Ink_solar cells, the capacitance value remains unaffected by such thickness variations.This suggests that the C1 value does not primarily reflect the polarization of the perovskite layer but is more likely related to an interface within the cell.It is worth noting that the current lumped model used was originally designed to fit the EIS spectra in planar device architectures.Considering the mesoporous structure of the studied device architecture, which involves a significant number of interface surfaces, it might be reasonable to contemplate adapting the model to accommodate this specific structural complexity.
Throughout the aging process, there is a noticeable contrast in behavior between drop-casting and inkjet-printing solar cells.For Ink_devices, a significant increase in parameter C1 is initially observed, and it remains then stable until reaching 1000 h of aging.For Drop_devices, this increase is more moderate, eventually reaching values like those achieved by ink cells ≈1000 h.This suggests that the electronic properties of the perovskite layer remain quite similar up to 1000 h of aging, which aligns with J(V) measurements showing similar V oc (open-circuit voltage) values at this stage of aging.However, the power conversion efficiency (PCE) values are lower for Ink_solar cells, which can likely be attributed to lower J sc (short-circuit current) values.This discrepancy in J sc may be explained by load transfer issues at the interfaces, affecting the overall performance of the Ink_solar cells.
On the contrary, C3 may exhibit much more variable values depending on the type of cells studied in the initial state.This aligns with the literature, which led the authors to associate C3 with the light-induced accumulation of an electronic accumulation zone formed by majority charge carriers (holes) near the cathodic contact (TiO 2 /Perovskite interface).Thus, we observe relatively similar values of C3 for both unmatured and matured Drop_solar cells, ≈5-6 mF.cm −2 , confirming previous observations made by the authors in a previous article, [40] indicating that for this type of solar cell, maturation does not lead to major changes at the interfaces.However, for Ink_solar cells, a significant increase in this parameter is observed in the initial state after maturation, confirming a significant effect at the interfaces during this stage.Subsequently, a general decrease in this parameter can be observed during aging, with this value remaining considerably lower for Ink_solar cells, indicating probable degradation in charge transfer at the interfaces, especially at the TiO 2 /perovskite interface.
In conclusion, these fruitful analyses explain why Ink_solar cells are much more sensitive to aging.On the scale of the active layer (R2/C1), a more premature degradation has been highlighted.Regarding charge transfers at the interfaces (R3/C3), a more significant degradation is also observed compared to Drop_solar cells.Furthermore, similar evolutions are noted regardless of the level of cell maturation, indicating no positive effect on the stability of such devices.
In conclusion, we propose an overview figure to highlight the correlation between the experimental results.Figure 11a presents a correlation between the measurements of power conversion efficiency (PCE) and short-circuit current (J sc ) based on the measured crystallinity ratio in the perovskite layer (using Xray diffraction), it seems that the variations in PCE and J sc are primarily due to changes in crystallinity ratio within the perovskite layer for Drop_solar cells.On the other hand, a completely different phenomenon is observed for Ink_solar cells, which show little evolution in their crystallinity ratio during aging.Even though all these cells exhibit relatively similar performances in the initial state, it is noticeable that the crystallinity ratios of the perovskite layers in Drop_solar cells are much higher.
Figure 11b presents a correlation between the measurements of photoluminescence intensity and recombination resistances measured by electrochemical impedance spectroscopy (EIS).While the photoluminescence intensity seems to show little variation with recombination resistances, an increase in photoluminescence intensity is noted with increasing recombination resistances, indicating that these recombinations are mainly radiative in nature and therefore not contributing effectively to the photovoltaic process.Since the values of R2 are similar for all cells (except at high aging times), these radiative recombinations predominantly occur at the interfaces.Consequently, it appears that the degradation mechanisms for the two types of cells studied are different, with the degradation linked to the perovskite layer for Drop_solar cells and to the interfaces for Ink_solar cells.These results appear to be consistent with the observations made in SEM.Indeed, as shown in Figure 11c,d), the active layer in ink_solar cells have a thinner thickness than in drop_solar cells.A thicker active layer promotes a better perovskite structure, while interface effects are amplified in the case of ink-deposited cells.

Conclusion and Perspectives
The objective of this article was to investigate the stability of Drop_ and Ink_solar cells under damp-heating conditions, focusing particularly on the effect of the maturation stage.Through a careful examination of the performance metrics, PV parameters, and observed "bumps" during J(V) measurements, it becomes evident that these phenomena are primarily attributed to the specific architecture (C-PSC) and perovskite characteristics (AVAI effect) employed in these cells.During a transient phase, charge transfers within the cell need to be rebalanced, and this phase is accelerated during the maturation stage performed immediately after the fabrication process or within the first 200 h of aging.Subsequently, it appears that the maturation stage has minimal impact, or at the very least, it may accelerate the degradation of the cells.
This study also highlights the surprisingly large influence of the perovskite deposition process, partly explained by the lower thickness in Ink_ compared to Drop_solar cells.Although the two types of cells performed similarly in the initial states (prior to and after maturation), their aging behaviors differ significantly.For Ink_solar cells, a gradual decline in performance is observed after an initial phase of mild aging until 200 h, eventually leading to complete cell degradation at 1000 h of aging.This performance decline is primarily attributed to a decrease in J sc and a significant increase in R s toward the end of the aging campaign.However, the V oc values of these cells remain much higher than those measured for Drop_solar cells throughout the aging period.In the case of Drop_solar cells, a decline in performance is observed up to 500 h of aging, followed by an improvement in performance up to 1000 h, reaching values comparable to the initial performance.However, a continuous decrease in performance is observed thereafter, until 3000 h of aging.The initial drop in performance is primarily driven by a reduction in V oc (while J sc remains relatively stable), reaching values similar to those obtained with inkjet cells.The subsequent decrease can be attributed to a significant drop in J sc .
The results obtained from the imaging techniques confirm the findings from J(V) measurements.Moreover, the LBIC measurements show much more homogeneous performances for the inkjet process.On the other hand, observations in PL imaging reveal much more homogeneous signals for the drop-casting process, partly explained by a thicker layer that responds more uniformly to the photoluminescence signal.
These results were further complemented by analyses of the perovskite layer conducted throughout the aging campaign, including photoluminescence spectroscopy, UV-vis, and X-ray diffraction (XRD).XRD analyses allowed the tracking of crystallinity ratios during aging.Prior to aging, it is noteworthy that the crystallinity ratio is significantly higher in Drop_solar cells, confirming that the perovskite structure is very different from that obtained in Ink_solar cells.Between 0 and 1000 h of aging, a reorganization of the crystalline structure of the perovskite occurs in drop-casting solar cells, accompanied by the formation and subsequent disappearance of amorphous and crystalline perovskite monohydrate.The mechanism leading to the second decrease in PCE is akin to that observed in inkjet-printed cells and follows a conventional pattern, mainly characterized by a decline in J sc due to a decrease in the perovskite's crystalline rate.These results are confirmed by the UV-vis spectroscopy data, which reveals a significant modification in the perovskite structure in Drop_solar cells between 0 and 500 h.Using this technique, it is also possible to track the formation of degradation products, such as PbI 2 .
To conclude the study, EIS measurements were conducted to further investigate the charge transfer mechanisms within the active layer and its interfaces.These measurements proved to be highly relevant as they not only confirmed the previous findings, particularly regarding much earlier and significant degradation in Ink_solar cells, but also revealed substantial degradation in charge transfer mechanisms at the interfaces, leading to a pronounced radiative recombination phenomenon, in agreement with the PL characterization.
In summary, although the maturation phase significantly enhances the initial performance of solar cells, its impact on device stability is not relevant.Conversely, the deposition method and the thickness of the active layer in these devices seem to play a crucial role in their stability.Even though the printing process allows better perovskite infiltration into the device, a lower quantity is infiltrated into the mesoporous network, resulting in a significant modification of the perovskite's microstructure.This exacerbates charge transport issues at the interfaces, which are also present while augmenting the perovskite quantity, revealing limitations occurring within the perovskite layer.
In the case of drop-casting cells, although their performance appears more promising and considerably more stable, developing large-area devices using this method seems challenging.However, these findings suggest that for potential industrialscale development of large-area devices, optimization of the printing process is necessary.This includes manipulating the number of injection points and their spacing to achieve stable and high-performing devices.
From a more fundamental perspective, the study conducted here highlighted the role played by the active layer/interface layers interfaces via the mesoporous network in the stability of carbon electrode-based cells.Given the architecture of the studied cells, interface effects between the active layer and interface layers are accentuated by the mesoporous network and hence substantial compared to planar cells.The positive aspect observed is the excellent stability of the perovskite in the bulk.Consequently, it would be intriguing to explore methods of enhancing these interfaces by modifying the active and interface layers or by incorporating additives.

Figure 1 .
Figure 1.a) C-PSC cell architecture with a simplified representation of the perovskite deposition processes: drop-casting and inkjet printing (active area = 12 × 12.5 mm 2 ); b) Dimension of solar cells after encapsulation (glass-glass laminated with a Surlyn gasket sealant, represented by a blue line).

Figure 2 .
Figure 2. Variations of PV parameters (Power Conversion Efficiency (PCE) presented respectively in normal and logarithmic scales a,b); Fill Factor (FF) c); Short Circuit Current density (J sc ) d); Open circuit voltage (V oc ) e); Series resistance (R s ) f)) during aging campaigns for "Drop-casting" (Drop) and "Inkjet-printing" (Ink) solar cells with (m) and without (w 0 ) the maturation pre-treatment step.(J(V) measurements were performed at 4 mV s −1 in reverse scan).

Figure 3 .
Figure 3. a) Variations of minimum and maximum J sc values (from J(V) measurements performed at different speeds) as a function of aging time for (Drop) and (Ink) solar cells with and without pre-maturation; J(V) measurements performed in reverse scan at different speeds as a function of aging time for unmatured (w 0 ) and matured (m) (Drop) (respectively b and c) and (Ink) (respectively d and e) solar cells.

Figure 4 .
Figure 4. a) LBIC maps for all studied cells for different aging times, the color scale represents the I sc variation; b) Variation of the LBIC measured average circuit current I sc during the aging campaign for drop and ink cells without (w 0 ) or with (w) pre-maturation (averaged over the active cell area, the uncertainty bar represents the variation over the area of one cell).

Figure 5 .
Figure 5. PL imaging photos for all studied cells for different aging times, the color scale represents the PL emission variation; b) Variation of photoluminescence measured by imaging during the aging campaign for drop and ink solar cells without (w 0 ) or with (w) pre-maturation (averaged over the active cell area, the uncertainty bar represents the variation over the area of one cell).

Figure 6 .
Figure 6.Photoluminescence spectra during aging for a) (Drop) (In inset, only spectra of matured devices Drop_m) and b) (Ink) solar cells; Variation of c) maximum emission intensity and d) Wavelength at maximum emission intensity  max as a function of aging time for (Drop) and (Ink) solar cells without (w 0 ) or with (m) maturation.

Figure 7 .
Figure 7. UV-vis reflectance spectra performed on complete cells (FTO side) during the aging campaign for a, b) drop and c) inkjet solar cells, d) Variation of reflectance value at 600 nm versus aging time for all solar cells studied.

Figure 9 .
Figure 9. Variation of XRD relative crystallinity of PK, PbI 2, and monohydrated perovskite versus aging time for a) drop and b) inkjet solar cells; Variation of PK orientation proportions during aging for c) drop and d) inkjet solar cells.

Figure 11 .
Figure 11.Correlation a) between PCE, Jsc and crystallinity rate of perovskite, b) PL intensity and R2+R3 evaluated by EIS measurements during aging campaign (up to 1000 h) for drop and ink solar cells without (w0) and with (m) maturation (zoom of the data in inset for small R2+R3 values); Active layer identification by SEM in c) drop_solar cell and in d) ink_solar cell.

Table 1 .
Attribution of main XRD diffraction lines.

Table 2 .
Statistical analysis (Mean: "Average calculated from the various measurements taken; SD: Standard deviation calculated from the uncertainties in the samples, with measurement and equipment uncertainties being much lower).
Yes (see Exp. Part) Photos/ mean ± SD = f(aging time) 2 samples EIS Yes (see Exp. Part) Key parameters extracted from Simulation/ representation as a function of aging time 2 samples