PbZrTiO3 Ferroelectric Oxide as an Electron Extraction Material for Stable Halide Perovskite Solar Cells

1. Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC and The Barcelona Institute of Science and Technology (BIST), Campus UAB, Bellaterra, E-08193 Barcelona, Spain. 2. Laboratory of Photonics and Interfaces, Institute of Chemical Sciences and Engineering, Ecole Polytechnique Fédérale de Lausanne, CH-1015 Lausanne, Switzerland. 3. Laboratory of Photomolecular Science, Institute of Chemical Sciences and Engineering, Ecole Polytechnique Fédérale de Lausanne, CH-1015 Lausanne, Switzerland.

Halide perovskite solar cells (PSCs) have emerged as a competitive photovoltaic technology with power conversion efficiencies (PCEs) surpassing the 22.7% mark. 1 One of the main bottlenecks of the technology is their long-term stability. 2 Understanding the different degradation mechanisms of the constituent materials, as well as interface instabilities, is of crucial importance for commercialization. 3 Semiconductor oxides (SO) constitute a fundamental part of highly efficient photovoltaic technologies 4-11 such as PSCs. 11,12 Oxides are applied as dense thin lms or as mesoporous oxide layers, working as hole or as electron transport materials (HTM or ETM, respectively) in normal or inverted PSC congurations. 10,11,[13][14][15] Electron transport semiconductor oxides, like TiO 2 , are characterized by an oxygen vacancy (O vac )-mediated conductivity 16 caused by a deviation in stoichiometry, the presence of impurities, or both. 17 In oxygen-containing atmospheres, and especially under UV light, holes generated at the nonstoichiometric oxide surface react with the oxygen adsorbed at O vac increasing charge recombination and degradation of the solar cell. 17,18 Different methods have been employed to passivate or eliminate these O vac . For example, the application of organic interfacial modiers with anchoring groups specically selected to bond with oxides, 3,19 the application of doped oxides 19,20 or the application of less reactive SnO 2 which results in less hygroscopicity, fewer O vac at its surface, and less UVdamage. 17,18 Another possibility is the application of a coating of secondary oxides, like ZrO 2 or Al 2 O 3, applied to suppress surface defects, avoid interfacial recombination, and enhance device stability. [21][22][23] More recently, Seok, et al., reported a highly stable PSCs by the application of an UV-light resistant oxide, the Ladoped BaSnO 3 , applied as electron transport layer. 24 A lessexplored option is the application of complex oxides with singular properties, such as ferroelectric, magnetic or ferroic oxides. Given their wide bandgap values and insulating properties, their application as electron transport materials (ETM) in emerging solar cells seems unlikely. 25 For example, ferroelectric oxides (FE-oxides) display spontaneous electric polarization where the application of an electric eld (poling) reorients domains within individual grains. Owing to this characteristic, different sources of conduction can be observed, such as polarization switching, breakdown and leakage. 26,27 Conduction through tunnel junctions 28,29 as in TiO 2 /Al 2 O 3 electrodes, 21 or domain walls 30,31 are other forms of current transport usually observed in nanometres thick lms. Moreover, their large dielectric constant implies the presence of a low exciton binding energy which translates in low electron/hole recombination and more defect-tolerant charge carrier transport. 32 Other ferroelectric materials have been applied in PSCs. For example ferroelectric polymers, such as PVDF, has been employed as an additive during the fabrication processes improving the perovskite morphology and the nal device efficiency. [33][34][35][36][37][38][39][40][41] It has also been applied as an insulating tunnelling layer between the halide perovskite and the electron transport layer, 42 increasing carrier transfer. 43 Nevertheless, the low photoactivity that FE oxides show under UV light makes them very attractive to enhance perovskite solar cell stability. In FE-oxides, UV light can photoinduce the enhancement of the remnant polarization, as recently reported for PbZrTiO 3 (PZT) lms. 44 The formation of defect dipoles, responsible for charge separation and current transport, depends on the presence of defect charges (O vac 3 . For comparison purposes, reference solar cells were made applying SnO 2 as dense ETM as in FTO/SnO 2 /CsMAFA/spiro-OMeTAD/ Au. We also explored the effect of the PZT layer on the triple cation perovskite under illumination with respect to solar cell stability. Analysis of the solar cells were made at 1-sun AM 1.5 G, including UV light, in air (no encapsulation), at 45 C and 55% RH. The solar cells were poled up to 2 V in order to polarize the PZT electrode. Poling resulted in a slow but steady improvement of the photovoltaic properties of the PZT-based PSC. The improvement was observed during the rst 90 minutes aer which the device stabilized and then maintained its photovoltaic properties under continuous illumination in air for hours. Our results demonstrate, for the rst time, the possibility of applying the PZT ferroelectric oxide as ETM in air and UV-stable PSCs.

Growth of crystalline PZT lms
In the context of emerging photovoltaics, ferroelectric oxide (FEoxide) layers with perovskite arrangement are of particular interest since they are a class of materials with unique functionalities, strong metal-oxide bonds, and high heat of formation resulting in very stable compounds. A highly crystalline structure enables the ferroelectric properties of the PZT, resulting in performance dependent on grain boundaries and the ferroelectric domain wall conguration. 44 Nevertheless, high temperatures, usually above 600 C, are required in general for the desired ABO 3 perovskite oxide phase formation 24 which limits the extrinsic conductivity of the transparent conducting FTO substrate. We circumvent this problem by the application of a fast (10 min) sintering method for the fabrication of PZT thin lms on transparent FTO. 11 The synthesis of the PZT, Pb(Zr x Ti 1Àx )O 3 (x ¼ 0.6) (for exact composition see ESI †), is made by solution processing and the solution is deposited on the FTO by spin coating. The minimum annealing temperature at which good ferroelectric properties are retained is found at 625 C. The ferroelectric perovskite crystalline structure of the PZT deposited on FTO was conrmed by X-ray diffraction (XRD) and scanning electron microscopy (SEM) analysis, as shown in Fig. 1 respectively (see also ESI Fig. S1 and S2 †). XRD theta-2theta (Fig. 1a) and cross-sectional X-TEM mapping (ESI Fig. S1 †) shows that the PZT thin lm belongs to the symmetry point group N m with space group Pm 3m (221) cubic, where the lattice constant has been determined to be a ¼ 4.077Å (ESI Fig. S1 †). Hysteresis loops demonstrating the ferroelectric character of the PZT lm were carried out to all thin lms. Fig. 3j and k shows the hysteresis loops of the PZT thin lm, synthesized at 625 C, before and aer irradiation with UV light (Fig. 3j) and visible light (k). The analyses were made rst in the dark and then aer exposed to different light sources: (nearultraviolet light-emitting diode (UVA-LED, 365 AE 10 nm/3.40 AE 0.12 eV) and white diode (White-LED, 400-700 nm/1.77-3.10 eV). For all the other PZT obtained at different temperatures see Fig. S2 (under ESI †). On the basis of controlled experiments examining the effect of the PZT thin lm thickness, a series of different PZT thickness were fabricated (see Fig. 1c and S4-S8 in ESI †). The screening of the thickness was accomplished by diluting the PZT solution in acetic acid (c/c o ) as described in methods (ESI Fig. S4 †). Analysis revealed that PZT lms ranged from thick 100 nm lms (corresponding to an undiluted solution c/c o of 1/0) to thin <10 nm lms (c/c o of 1/10). Two dissimilar samples were then selected, a thick 100 nm and a thin $10 nm lm ( Fig. 1a and b). The nal PZT surface morphology, analysed by AFM and SEM ( Fig. 1e-h), seems to be poorly affected by the lm thickness but also likely inuenced by the well-crystallized FTO substrate which has a rutile crystal structure with a tetragonal unit cell (P4 2 /mnm) matching the perovskite oxide structure. Nevertheless, careful analysis of the surface of a series of different PZT thin lms made at different thicknesses showed a slight increase in surface roughness when increasing thin lm thickness (ESI Fig. S5 †). The surface morphology of the thicker lms (e.g. 100 nm) are free of microcracks, pores or holes, whereas some pinholes can be present in thinner thin lms.

Photovoltaic performance of the PZT-based PSCs
Here we fabricate complete PSCs consisting of the FTO/PZT thin lms, followed by the CsMAFA perovskite of about 600 nm. Spiro-OMeTAD was used as the hole transport material and a thin layer (80 nm) of Au as the back metal electrode. For comparison purposes, reference solar cells were also fabricated as previously described 45    (e-f) AFM surface analysis of the $10 nm thin PZT film; (g and h) surface SEM images (g) and grain boundaries (marked in white, (h)) of the $10 nm thin PZT film. were then subjected to poling up to 2 V under 1 sun AM 1.5 G. During poling, the photovoltaic properties of the PZT-based solar cells increased gradually until stabilization. This improvement was almost negligible for the thick 100 nm PZT based PSC (ESI Fig. S7 †) and drastic for the thin $10 nm PZTbased PSC (Fig. 2c-f). The champion PSCs corresponds to solar cells with the $10 nm PZT thin lm ( Fig. 2c-f) which showed a dramatic and steady increase in PCE scaling from an initial 1.49% as-prepared, before poling ( Fig. 2e), up to 11.16% efficiency aer poling (Fig. 2f, Table 1). Fig. 2d depicts the histogram of the PCE for the PSCs using the champion cell of $10 nm thick PZT lm, before and aer poling, displaying a very reproducible pattern of efficiency increase for all devices. Besides the difference in thicknesses, the dissimilarity observed on photovoltaic response between the solar cells with the thin ($10 nm) and the thick (100 nm) PZT lm can be related to the surface roughness morphology and grain size of the PZT layer. The polarization-dependent conductivity (ferroelectric properties) 48,49 is highly reliant on synthesis conditions, morphology, defect concentration and the orientation and size of the oxide grains. 50 Especially the presence of different grain size indicates different intrinsic stress in the lms due to the difference in grain boundaries, resulting in different dielectric properties. 51 Thus, the different thickness, surface roughness morphology and grain size of the PZT lms could benet the photovoltaic response aer poling for the PSC with the $10 nm thin PZT which shows the largest roughness (Ra $26 nm) in comparison to the 100 nm thick PZT (Ra $ 20 nm) (ESI Fig. S2 †).
Electrical poling under 1 sun AM 1.5 G Poling, the process of applying an external electric eld to a ferroelectric material, induces the reorientation of dipole moments within individual grains improving the mobility of charge carriers. 52 Poling conditions carried out under 1 sun AM 1.5 G, were observed to be strongly affected by the PZT thin lm thickness, the light harvester material (active layer), and the method at which the electric bias is applied. In our samples, progressive IV poling (with increasing initial voltage of 20-50 mV) were observed to result in enhanced photovoltaic performance of the FE-oxide based PSC if compared to the application of a single high voltage. Thus, careful optimization of all parameters was carried out (see ESI Fig. S2, S8-S11 †). We rst analysed the voltage limit for poling. We have previously observed high robust response of solar cells based on the P3HT:PCBM active layer even aer the application of voltages as high as 40 V. 53 In the case of the halide perovskite active layer, the $10 nm PZT-based PSCs barely sustain the application of 2 V, while the thick 100 nm PZT-based PSC requires stronger poling showing a voltage limit at around 4-5 V. Moreover, the scan rate should also be limited to low speeds, with optimum values below 2 mV s À1 . Finally, we observed that the solar cell was unresponsive aer several scans in the forward direction (i.e. biasing the solar cell from the minimum positive voltage value) during poling, thus poling was carried out under reverse mode only. In summary, the poling was carried out in air at under 1 sun AM 1.5 G (including UV light), through subsequent IV-curves up to maximum of +2 V. The scan direction was the reverse mode (i.e. biasing the solar cell from the maximum positive voltage value) and the scan rate was 2 mV s À1 . IV curves were performed every 5 min until the photovoltaic parameters, in particular current density, J sc , reached a maximum and stabilized (for details on poling optimization see ESI Fig. S7, S8 and S15 †). Fig. 3a and b shows the effect of electrical poling on the photovoltaic response of the champion PZT-based PSC (red line) and the comparison with a reference device based on SnO 2 (blue line). The PZT-based PSC increased its photovoltaic response from 1.49% up to 11.16% aer poling. This response validates the effective activation of the photovoltaic parameters through poling for the PZT which evolve slowly until stabilization of the PCE aer 90 minutes. Poling the reference solar cell (made with SnO 2 ) results in a clear degradation of the photovoltaic response aer poling (Fig. 3b). Although this observed degradation of the SnO 2 -based PSC is expected (due to the severe testing environment and conditions) it permits to demonstrate the dissimilar working mechanism between both the PZT and the SnO 2 ETMs. The poling effect is observed to be highly reproducible for a series of analysed samples as shown by the error bars depicted in Fig. 3c-f (also in Fig. S7 †). These results demonstrate the crucial importance of poling under light irradiation, for the successful transformation of the insulating PZT into an electron transport material in our solar cells.
To analyse the evolution of the transport properties of the PZT-based PSC under UV light and poling, we carried out experiments of IPCE (or EQE) against poling time. We then compared the J sc obtained from the IPCE and the one obtained from the IV curves at specic times. The poling experiments were carried out at 0.1 h, 1.0 h, 2.5 h and 5 h following the same Table 1 Photovoltaic parameters of the champion PSC applying $10 nm thin film PZT-based PSC and the reference PSC applying SnO 2 before and after poling. Values are given for the forward and reverse scans. Solar cells were analysed at 1-sun AM 1.5 G, including UV light, in air (no encapsulation), at 45  poling procedure. Aer each poling period the sample was taken out of the sun simulator and immediately analysed by IPCE. The rst observation is that, due to the insulating properties of the PZT, very low charge transport is initially observed (Fig. 3g-1) by increasing the poling time from 0.1 h to 5 h, the conversion of photons to electrons increases from 4% to 60% (Fig. 3g-1 and 4). The transformation of the PZT with time and the improvement of IPCE (and solar cell efficiency) can also be explained by the shape of the IPCE spectra. Fig. 3h shows the normalized IPCE spectra shown in Fig. 3g. The initial spectra obtained aer 0.1 h of poling ( Fig. 3h-1) revealed a very weak absorption of the PZT oxide at wavelengths below 340 nm, corresponding with a very poor semiconductor. At this time, no absorption of the halide perovskite CsMAFA (between 400-800 nm) is observed since no charge transport takes place from the PZT to the halide perovskite. As the poling under UV light continues, this peak at 340 nm, red shis to 380 nm (from 3.64 to 3.26 eV), a response observed during the reduction of the band gap in oxides. This red shi is attributed to oxide doping observed in classical oxides like TiO 2 (ref. 54 and 55) but also observed in other oxides like PZT. 56 It has also been attributed to the continuous and steady removal of O vac under UV light exposure or inert atmospheres for several binary oxides. 6,7 The increase of the poling time, from 0.1 h to 1.0 h and 2.5 h, results in the gradual improvement of the IPCE absorption between 400 nm and 800 nm, which is the region that corresponds to the absorption of the halide perovskite CsMAFA material (Fig. 3h-1  to 4), an indication of charge transport throughout the PSC. We should also emphasize that the shape of the nal IPCE obtained aer 5 h poling (Fig. 3g-4 and h-4) corresponds exactly with the shape of the IPCE of a PSC applying the classical SnO 2 oxide as ETL as shown in ESI Fig. S9 †. 57 Finally, we compared the response observed for the current density (J sc ) obtained from IV curves and IPCE aer poling (Fig. 3i). In both cases, the J sc shows a steadily increase until reaching similar values around $14 mA cm À2 aer the poling process. We observed however, that for intermediate poling times between (1) 0.1 h, (2) 1.0 h and (3) 2.5 h, the J sc obtained from IPCE is always lower than the J sc obtained from the IV curves (Fig. 3i). The main difference between both characterization methods is that IV curves are carried out under full illumination under the sun simulator, right aer poling. In the case of the IPCE characterization, illumination and poling are eliminated in order to situate the sample under the IPCE set up for analysis. In this short period of time, the removal of UV light and poling in PZT samples brings the material back to its original state and electrostatic repulsion between polarization charges and defect charges (O vac ) immediate occurs, depolarizing the PZT lm. Since remnant polarization is dened as the polarization remaining in the material aer the poling voltage is eliminated or reduced to zero, we attributed the increase in J sc (from IPCE) to the enhancement of the remnant polarization when increasing poling time, in perfect agreement with the mechanism reported for PZT by Cao, et al. 44 Thus, we infer here that the as-prepared PZT-based PSC with polycrystalline PZT thin lms (as shown in Fig. 1 and ESI  Fig. S6 †) have domains which are random. When the solar cell is poled under UV light, part of the defect-dipoles located at O vac , are separated and the charges are driven to opposite interfaces of the lm. This effect is enhanced by UV light irradiation which permits the excitation of the remaining defectdipole charges that are expelled to opposite interfaces by the poling eld as reported for PZT. 44 The process takes place slowly throughout the rst minutes of the analysis. With time, the PZT behaves as an electron transport material in the solar cell and delivers the best photovoltaic performance. The effect is reversible if the bias voltage and UV light is eliminated but can become permanent if the poling time increases (as observed in Fig. 3i), in agreement to the enhancement of the remnant polarization with poling as described for PZT. 44 The transformation of the PZT from an insulator into a conductor is proved by the hysteresis (ferroelectric) loops analysed to the PZT before and aer light irradiation ( Fig. 3j and k). Aer analysis under white-LED the hysteresis loop shows a clear ferroelectric response. Under UV-LED irradiation the PZT samples presented signicant hysteresis suppression in the hysteresis loops as observed in Fig. 3j. This effect, known as polarization suppression 58 (or ferroelectric fatigue 59 ), is a well-known effect in ferroelectric materials 60 and it is directly related to the conductivity of the perovskites: good ferroelectric properties can only be observed on materials with low conductivity, improving conductivity will inevitably increase the leakage current and suppress ferroelectricity. 60 These results are also supported by conductivity measurements carried out to the PZT thin lm in the dark and aer light irradiation (Fig. 3g). The PZT samples analysed in the dark show low conductivity which increases once the sample is irradiated with light. The effect is reversible once the PZT thin lm is stored in the dark again.
Given that the defect-dipoles are reliant on the multicrystalline nature of the PZT (density of grain boundaries and domain walls), we infer that the different response of the PSC aer poling is attributed to the different PZT thickness (and surface roughness) as shown in Fig. 1 (and Fig. S6 ESI †). We also observed that the PZT-based PSCs showed large hysteresis aer poling and that the hysteresis index (HI) strongly depends on the poling time ( Fig. 2 and ESI Fig. S14 †). Since hysteresis is related to factors like applied conditioning voltage, device conguration or UV light, among many others 61 we attribute the observed hysteresis to the effect of testing conditions (ambient air or the presence of UV light) but especially to the large poling voltage applied. Especially, the hysteresis observed from ion migration in halide perovskites can be strongly affected by the poling procedure. Identifying a specic conduction mechanism in a device comprising a ferroelectric material can be rather complex, especially in a dynamic system under the continuous effect of poling and UV light. Nevertheless, since several possible conduction mechanisms can be possible (for example bulk effects including ohmic conduction, charge carrier hopping, space charge limited conduction, interface effects, tunnelling, conduction through domain walls and the conduction mechanisms related to defects like oxygen vacancies in oxides 58-60 ), there is a possibility that poling and UV light irradiation can also be causing the accumulation of charges or the migration of ions (or the combination of all of them) in the active, transport layer or at interfaces in our devices.

Stability analysis in air and UV light
We further studied the stability of PSCs under continuous 1-sun AM 1.5 G light irradiation (Fig. 4). Aer poling, the PSCs were analysed by consecutive IV curves taken each few minutes in reverse mode with a scan rate of 2 mV s À1 . The I-V scan starting voltage was established at +1.6 V and the scan nal voltage was À0.1 V. Degradation of halide PSCs under ambient conditions (without encapsulation) is well-documented for devices applying electron transport oxides like TiO 2 or SnO 2 . 3,11,14 Thus, long-term stability studies are usually carried out under encapsulation conditions or inert atmospheres. 24 In our stability tests, we want to demonstrate the effective response of the PZT thin lm under severe conditions: 1 sun continuous light irradiation (including UV light), in air, at 45 C and 55% RH and without encapsulation. Fig. 4 shows the results of the initial 10 h of the stability analysis carried out aer poling. Strikingly, our PZT-based PSC shows a remarkable stability under air and UV light, while the SnO 2 -based PSC degrades in minutes. Although one can argue that the degradation of the reference device, the SnO 2 -based PSC, is most likely to be intensied due to the application of bias voltage (poling), the fact that our PZT-based PSCs sustain and improve its photovoltaic properties demonstrates a different working mechanism and an improved stability in air and UV light.

Conclusions
We introduced the ferroelectric oxide Pb(Zr x Ti 1Àx )O 3 (x ¼ 0.6), PZT, as the electron extraction layer in triple cation, This improvement is directly inuenced by the thickness of the PZT with the best response observed for PZT lms of $10 nm. The champion PZT-based solar cell resulted in 11% PCE aer poling under standard 1-sun AM 1.5 G sunlight (including UV light, in air and without encapsulation). The solar cells were stable in air and UV light for several hours demonstrating the stability of the CsMAFA perovskite. We also demonstrated a different working mechanism of the PZT in comparison with a classical oxides like SnO 2 or TiO 2 , establishing for the rst time the possibility to apply (and transform) an insulating ferroelectric oxide to work as electron extraction layer in PSCs.
Our experiments indicate the effect of ferroelectricity, however alternative conducting mechanisms affected by the accumulation of charges or the migration of ions (or the combination of them) cannot be ruled out. We expect our ndings to have a major impact in a new generation of ETMs for PSCs that can also affect the long-term stability of emerging photovoltaics, but also to open up new possible functionalities of ferroelectric oxides in a wide range of novel electronic and optoelectronic applications.

Conflicts of interest
There are no conicts to declare.