Two birds with one stone: dual grain-boundary and interface passivation enables 4 22% eﬃcient inverted methylammonium-free perovskite solar cells

Advancing inverted (p–i–n) perovskite solar cells (PSCs) is key to further enhance the power conversion efficiency (PCE) and stability of flexible and perovskite-based tandem photovoltaics. Yet, the presence of defects at grain boundaries and in particular interfacial recombination at the perovskite/electron transporting layer interface induce severe non-radiative recombination losses, limiting the open-circuit voltage ( V OC ) and fill factor (FF) of PSCs in this architecture. In this work, we introduce a dual passivation strategy using the long chain alkylammonium salt phenethylammonium chloride (PEACl) both as an additive and for surface treatment to simultaneously passivate the grain boundaries and the perovskite/ C 60 interface. Using [2-(9 H -carbazol-9-yl)ethyl]phosphonic acid (2PACz) as a hole transporting layer and a methylammonium (MA)-free Cs 0.18 FA 0.82 PbI 3 perovskite absorber with a bandgap of B 1.57 eV, prolonged charge carrier lifetime and an on average 63 meV enhanced internal quasi-Fermi level splitting are achieved upon dual passivation compared to reference p–i–n PSCs. Thereby, we achieve one of the highest PCEs for p–i–n PSCs of 22.7% (stabilized at 22.3%) by advancing simultaneously the V OC and FF up to 1.162 V and 83.2%, respectively. Using a variety of experimental techniques, we attribute the positive effects to the formation of a heterogeneous 2D Ruddlesden–Popper (PEA) 2 (Cs 1 (cid:2) x FA x ) n (cid:2) 1 Pb n (I 1 (cid:2) y Cl y ) 3 n +1 phase at the grain boundaries and surface of the perovskite films. At the same time, the activation energy for ion migration is significantly increased, resulting in enhanced stability of the PSCs under light, humidity, and thermal stress. The presented dual passivation strategy highlights the importance of defect management both in the grain boundaries and the surface of the perovskite absorber layer using a proper passivation material to achieve both highly efficient and stable inverted p–i–n PSCs. Due to the rapid increase in power conversion eﬃciency of perovskite solar cells (PSCs), they are considered an emerging area of research in photovoltaic technologies. While inverted p–i–n PSCs have demonstrated great potential for flexible and perovskite-based tandem photovoltaics, key challenges still need to be addressed as compared to their n–i–p counterparts. In particular, severe non-radiative recombination losses induced by the presence of defects at grain boundaries (GBs) and interfacial recombination at the perovskite/electron transporting layer interface limit the open-circuit voltage ( V OC ) and fill factor (FF) of PSCs in this architecture. To address this issue, we demonstrate that utilizing a dual passivation strategy using phenethylammonium chloride both as an additive and for surface treatment simultaneously passivates defects at the GBs and the perovskite/C 60 interface. We show that this is due to the formation of a heterogeneous 2D Ruddlesden-Popper phase, leading to a significant improvement in both the V OC and FF. In view of the urge to advance p–i–n PSCs for flexible and perovskite-based tandem photovoltaics, our findings stress the importance of defect management both at the GBs and the surface of the perovskite absorber layer in order to achieve both highly efficient and stable inverted p–i–n PSCs.

Post-treatment of perovskite films is a widely established strategy to suppress interfacial recombination and optimize the performance of PSCs. 49,[52][53][54][55][56][57][58] Prominent examples for tailored passivation schemes are the use of (alkyl)ammonium salts, 3,4,6,10,11,13,14,20,22, other organic compounds 16,19,24,45,[81][82][83][84][85][86][87] and fluoride-containing materials. 31,35,38,74,86,88,89 Alongside established chemical passivation that reduces the density of surface/interface defects, 53,63,83,85 this strategy also encompasses performance enhancements by the formation of 2D/3D heterostructures 10,[59][60][61]63,64,70,[76][77][78] and/or wide-bandgap interface layers. 6,66,90,91 The latter enhancements can be the result of improved energy level alignment that promotes selectivity and carrier transport across perovskite/CTL interfaces and/or a reduced probability for interfacial recombination due to charge blocking. 19,25,35,53,67,77,92 Recently, lithium fluoride (LiF) has been identified as an interlayer at the perovskite/ETL interface that significantly enhances the performance of p-i-n PSCs. 31,35,38,88,89,93 However, PSCs with LiF undergo severe long-term degradation which limits the applicability of this approach. 31,93 In order to reduce non-radiative recombination in the bulk and grain boundaries, the use of non-stoichiometric precursors [94][95][96] or incorporation of different additives into the perovskite precursor solution or antisolvent such as metal cations, 97 anions, 98 chloride (Cl) or thiocyanate (SCN), [4][5][6]10,14,15,17,19,88 (alkyl)ammonium salts, 3,25,26,64,[98][99][100][101][102][103][104][105] other organic compounds, 74,84,106 and fluoride-containing materials 74,107 have been proposed. Given that these additives directly assist in perovskite film formation, changes in crystallization dynamics as well as a reduced defect density are commonly observed. 55,108,109 For instance, Xu et al. demonstrated that by alloying MAPbCl 3 into the perovskite film, the V OC of the wide-bandgap p-i-n PSCs significantly improved due to a reduced bulk defect density. 88 In other works, the addition of various long chain alkylammonium cations was shown to self-assemble into a wide-bandgap 2D perovskite phase passivating the surface and/or grain boundaries of the 3D perovskite film. 25,58,64,[99][100][101][102][103][104][105] However, due to the insulating nature of such 2D phases, adding too large amounts typically results in an overall lower PCE compared to control devices. 25,64,99,101,103,104,107 Despite the apparent wide range of strategies suggested to reduce non-radiative recombination losses in PSCs, 49,54,57,110,111 to our knowledge there is only one recent report that used the same passivation material both as an additive and for surface treatment to improve the performance of p-i-n PSCs. 84 In this work, we report on an effective dual passivation approach using the long chain alkylammonium salt phenethylammonium chloride (PEACl) to simultaneously passivate the grain boundaries and the perovskite/C 60 interface by using PEACl:PbCl 2 as the additive and PEACl for surface treatment, respectively. Employing time-resolved photoluminescence (TRPL) and photoluminescence quantum yield (PLQY) measurements, dual passivation is proven to be most effective in reducing nonradiative recombination compared to either of the individual passivation strategies. By analyzing cathodoluminescence (CL), scanning electron microscopy (SEM), X-ray/ultraviolet photoelectron spectroscopy (XPS/UPS), X-ray diffraction (XRD) and Kelvin probe force microscopy (KPFM) measurements, we attribute the positive effects to the formation of a heterogeneous 2D Ruddlesden-Popper (RP) (PEA) 2 (Cs 1Àx FA x ) nÀ1 Pb n (I 1Ày Cl y ) 3n+1 perovskite phase with n B 1-2 at the surface and grain boundaries of the film that exhibits a lower work function (WF) and hole blocking properties. Finally, thermal admittance spectroscopy (TAS) reveals that the activation energy for ion migration is strongly increased upon dual passivation, which is reflected in an enhanced device stability under maximum power point (MPP) tracking and heat treatment for 1000 h. In summary, by using PEACl both as the additive and for surface treatment, we could not only effectively reduce interfacial recombination at the perovskite/C 60 interface, but simultaneously passivate the grain boundary defects. Employing dual passivation for methylammonium (MA)-free p-i-n PSCs with a bandgap of B1.57 eV leads to a very high PCE of 22.7% (stabilized at 22.3%) with a remarkable V OC and FF of up to 1.162 V and 83.2%, respectively. In view of the urge to advance the p-i-n structure for flexible and perovskite-based tandem photovoltaics, this development is pivotal.

Results and discussion
The dual passivation strategy developed in this work is based on combining the incorporation of PEACl:PbCl 2 into the perovskite precursor solution and PEACl surface treatment (Fig. 1). Since using long chain alkylammonium salts as additive mainly passivates the grain boundaries, as will be shown later and has been proposed in previous works, 26,98,[101][102][103] it is for simplicity referred to as grain boundary passivation (GBP) in the following, while surface treatment is referred to as surface passivation (SP  (Fig. 1a). In case of SP, PEACl dissolved in isopropanol (optimized at a concentration of 1.5 mg ml À1 ; see films and subsequently annealed at 100 1C for 10 min (Fig. 1b). Finally, both individual passivation strategies are exploited together, referred to as GBP&SP (Fig. 1c). Further details are provided in the Experimental Section (ESI †).

Photovoltaic performance
In order to demonstrate the trend in performance of planar p-i-n PSCs upon employing either individual passivation (GBP or SP) as well as dual passivation (GBP&SP) compared to Ref PSCs, devices in the layer stack ITO/2PACz/perovskite/C 60 /BCP/ Ag with an active area of 12.3 mm 2 were prepared (Fig. 2a). Fig. 2b and Table S1 (ESI †) summarize the current densityvoltage ( J-V) characteristics and PV parameters of the bestperforming p-i-n PSCs. The corresponding statistics (in total 147 devices) that emphasize the very high yield and good reproducibility of the key trends are shown in Fig. 2c. The best Ref PSC exhibits a PCE of 20.4% with a short-circuit current density ( J SC ) of 23.9 mA cm À2 , a V OC of 1.086 V, and a FF of 78.6%. This denotes a very respectable starting point in performance for p-i-n PSCs compared to literature (see Fig. S1, ESI †). The PCE of the best GBP PSC is slightly improved to 20.7%, which is mainly associated with a 26 mV enhancement in V OC as well as a slightly improved FF. The small decrease in J SC to 23.6 mA cm À2 is attributed to the formation of a 2D RP perovskite (see discussion in the following) and, thus, a slight decrease in 3D perovskite absorber volume (see ultravioletvisible (UV-vis) measurements in Fig. S5, ESI †). 64,99 The best SP PSC already exhibits a very high PCE of 22.1% with a significant improvement in both the V OC (1.131 V) and FF (82.3%) as compared to the Ref PSC.
Strikingly, upon dual passivation, the V OC and FF are further enhanced to 1.162 V and 83.2% respectively, which leads to a remarkable PCE of 22.7% for the best GBP&SP PSC (see Fig. 2d). This corresponds to a V OC Â FF product of 0.891 with respect to the S-Q limit, the highest reported for p-i-n PSCs with a PCE above 21% (compare Fig. S1 and S2b, ESI †). Furthermore, the GBP&SP PSC also exhibits a remarkable stabilized PCE (under MPP tracking), V OC and J SC of 22.3%, 1.161 V and 23.4 mA cm À2 under continuous AM1.5G illumination for 5 min, respectively (Fig. 2e). It should be noted that the V OC enhancements are not governed by an increase in the bandgap, as shown by analysis via the Tauc plot method and the inflection point of the EQE spectra ( Fig. S6a and b, ESI †), 113 but relate to reduced nonradiative recombination, as we will later elaborate on in detail. We note that when increasing the concentration of PEACl to 3 mg ml À1 (beyond the optimum concentration), the V OC of GBP&SP PSCs increases further up to 1.184 V (see Fig. S4, ESI †), which represents a voltage deficit of only 393 mV and 104 mV with respect to the bandgap and radiative limit respectively (90.9% of the S-Q limit), 113 that are among the lowest reported for p-i-n PSCs (Fig. S2a, ESI †). However, since the J SC and FF decline at the same time, possibly due to the insulating nature of a thicker 2D RP passivation layer at the surface, 63,64,81,82,99 the PCE of the best GBP&SP PSC drops to 21.8%. It should be noted that the reported J SC for all PSCs is corrected using the ratio of J SC derived from the external quantum efficiency (EQE) and J-V measurements of the best PSCs ( Fig. 2b and Fig. S7, ESI †). This results in a rather conservative determination of PCE in this work. To summarize, the combined enhancements in V OC and FF are highest for GBP&SP PSCs, which highlights the necessity of the simultaneous passivation of the perovskite/C 60 interface, and -as we will show later -the grain boundaries of the perovskite thin film.
Next to dual passivation by PEACl, we first evaluated the effect on the V OC upon employing PEAI and PEABr, since they have been used in numerous previous reports for passivation of perovskite films. 62,65,69,77,99,103 As shown in Fig. S8 (ESI †), PEACl-based GBP&SP PSCs show a much higher average of V OC of B1.15 V as compared to B1.12 V in case of PEAI and PEABr. Therefore, we focussed in more detail on alternative chloride-based long chain alkylammonium salts namely nbutylammonium chloride (BACl) and n-octylammonium chloride (OACl), since BAI, 63,64,71,72,76,102 BABr, 10,59,60,64 OAI,4,72,75,76,101 and OABr 10,71 have previously been reported to serve as efficient passivation molecules as additive as well as for surface treatment.
While SP PSCs all exhibit an enhanced V OC and FF as compared to Ref PSCs, the enhancements are most pronounced in case of PEACl (Fig. S9, ESI †). Employing the dual passivation strategy leads to aB30 mV V OC enhancement in case of OACl and BACl, which is much lower compared to B70 mV for PEACl-based GBP&SP PSCs (see Fig. S10, ESI †). Furthermore, while for BACl-based GBP&SP PSCs the FF remains similar and only a slight drop in J SC is observed compared to Ref PSCs, these parameters are even reduced in case of OACl, as expected based on previous reports employing too large amounts of alkylammonium salts as additive. 25,64,99,102 Therefore, an overall lower average PCE of only 18.6% is obtained for OACl-based GBP&SP PSCs compared to 20.1% for Ref PSCs, while the average PCE is slightly higher at 20.9% in case of BACl. These results highlight that our dual GBP&SP passivation strategy in principle is compatible with other Cl-based long chain alkylammonium salts, but reduced charge carrier transport (i.e., lower J SC and/or FF) can easily impede any positive effects from reduced interfacial recombination (i.e., higher V OC ). Hence, careful optimization of the fabrication parameters is required. Targeting high efficiency and reproducibility, we identified PEACl as the superior choice for the dual passivation strategy of p-i-n PSCs studied in this work.

Photophysical properties
To discriminate the effect of GBP, SP and GBP&SP on nonradiative recombination of the PSCs, we first show representative TRPL transients measured for ITO/2PACz/perovskite/C 60 layer stacks in Fig. 3a. While the detailed interpretation of such transients can be challenging, 114 a longer monomolecular

View Article Online
lifetime at low-level injection can be attributed to reduced nonradiative recombination either within the bulk (including the grain boundaries) or the perovskite/CTL interfaces. [114][115][116][117] The lifetime increases considerably by more than one order of magnitude in the order ref (19 ns) -GBP (48 ns) -SP (113 ns) -GBP&SP (256 ns) (dashed lines Fig. 3a). This indicates that non-radiative recombination is effectively suppressed in the same order as the observed V OC enhancement of the PSCs. [114][115][116][117] To quantify the reduction of non-radiative recombination, PLQY measurements along with the internal quasi-Fermi level splitting (E F ), that are attributed to the 'implied V OC ' via V OC_imp = DE F /q = V OC_rad + k B T/q ln(PLQY), are discussed next. 34,118 Analysing PLQY and V OC_imp for the stack ITO/2PACz/perovskite without C 60 allows identification of whether non-radiative recombination at the HTL/perovskite limits the V OC of our PSCs. 34,35,39 For the Ref films, we find an already very high average PLQY (V OC_imp ) of 7.2% (1.206 V) which only slightly increases to 7.9% (1.218 V), 9.8% (1.218 V) and 9.7% (1.225 V) for GBP, SP and GBP&SP films, respectively (Fig. S11, ESI †), with V OC_imp being well above the obtained V OC of all PSCs presented in Fig. 2c. This shows that our perovskite films are of very high quality and that the nearly lossless 2PACz/perovskite interface does not limit the V OC , in line with previous reports. 31,32,35 Upon addition of C 60 /BCP/Ag, the PLQY for Ref films severely drops to a low average value of 0.058% correlating to V OC_imp of 1.081 V, clearly showing that the perovskite/C 60 interface limits the V OC ( Fig. 2b and c). Impressively, the average PLQY increases by roughly one order of magnitude to 0.083%, 0.26% and 0.45% for GBP, SP, and GBP&SP films respectively (Fig. 3b), correlating to an enhanced V OC_imp of 1.100 V, 1.122 V and 1.144 V, respectively (Fig. 3c). Notably, the values of V OC_imp closely match with the average V OC of the respective PSCs (compare Fig. 2c), implying that all PSCs have a proper energetic alignment that does not result in an offset between V OC_imp and V OC . 34,35,92,118,119 We note that increasing the PEACl concentration to 3 mg ml À1 for GBP&SP films further increases the PLQY and V OC_imp to remarkable values of up to 2.21% and 1.190 V, respectively, which is in line with the results discussed above for the respective PSCs (see Fig. S12, ESI †).
At first sight, the role of the PEACl:PbCl 2 additive for the improved device performance remains unclear, since V OC_imp for the half layer stack without C 60 only slightly increases for all passivation strategies as compared to Ref films. To shed more light on this aspect, we show representative TRPL transients for the stack ITO/2PACz/perovskite in Fig. S13 (ESI †). Here, we find a clear trend with the monomolecular lifetime for GBP films (B1624 ns) and GBP&SP films (B1497 ns) being considerably longer as compared to Ref films (B335 ns). Interestingly, for perovskite films with surface passivation only, the lifetime is solely slightly increased to B464 ns. At this point, we hypothesize that the enhanced lifetimes in case of GBP relate to the passivation of shallow grain boundary traps via the selfassembly of PEA + molecules and/or the formation of a PEAClbased 2D RP phase. 25,26,36,41,46,102,[104][105][106]120,121 We note that shallow traps are typically filled at high illumination intensities around 1 Sun, possibly explaining why the values of PLQY and V OC_imp for stacks without C 60 are only slightly enhanced by all three passivation strategies. This explanation is in line with the common implication that grain boundaries are not necessarily detrimental to device performance at solar illumination intensities. 40,41,120 To shed more light on this, we evaluate the trap-state density (n t ) and charge carrier mobility (m) of electrons and holes for the different passivation strategies via space charge limited current (SCLC) measurements. We fabricated both electron-and hole-only devices with the configuration of ITO/ SnO 2 /perovskite/C 60 /BCP/Ag and ITO/2PACz/perovskite/Spiro-MeOTAD/Ag, respectively. The dark J-V characteristics of the devices are plotted in Fig. S14 and S15 (ESI †) and are analyzed according to the SCLC method (see further details in the ESI †). The electron mobility of the Ref device is 4.9 Â 10 À3 cm 2 V À1 s À1 , while both SP and GBP devices demonstrate a comparable increase in mobility to 6.3 Â 10 À3 and 7.2 Â 10 À3 cm 2 V À1 s À1 (see Table S2, ESI †), respectively. Applying our dual passivation strategy, the electron mobility further increases to 10.0 Â 10 À3 cm 2 V À1 s À1 . Furthermore, the trap-filled limit voltage (V TFL ), which is linearly proportional to the trap-state density, demonstrates a substantial decrease in the order Ref -GBP -SP -GBP&SP, correlating to a reduction in electron trap density from 9.2 Â 10 15 cm À3 to 6.5 Â 10 15 , 5.4 Â 10 15 and 3.7 Â 10 15 cm À3 , respectively (see Fig. S14 and Table S2, ESI †). A very similar trend is also observed for the calculated hole mobilities, while the reduction in the hole trap density is apparently slightly less pronounced (see Fig. S15 and Table S3, ESI †). Therefore, the TRPL and SCLC results indicate that both GBP and SP independently contribute to enhancing both the electron-and hole mobilities by roughly a factor of 2, while at the same time specifically reducing the electron trap density at the grain boundaries and surface of the perovskite film.
To assess the impact of the reduced trap-state density on device performance, we perform intensity-dependent PLQY measurements to obtain the internal ideality factor (n id ) from a fit to the calculated V OC_imp . 34 The ideality factor has been proven to be governed by bulk as well as interfacial recombination properties. 34,122,123 For high-performing PSCs that are not limited by failures at either of the perovskite/CTL interfaces, a reduction of n id towards a value of 1 is typically associated with a predominant bimolecular radiative and reduced trap-assisted Shockley-Read Hall recombination and there is a direct correlation between n id and FF. 31,34,122,124 Again, we first analyse half-layer stacks without C 60 and find a considerable reduction of n id from 1.71 for Ref films to 1.48 and 1.60 for GBP and SP films, respectively, while GBP&SP films show by far the lowest n id of 1.30 (Fig. S16, ESI †). The observation that n id in case of GBP is slightly lower as compared to SP is in line with the enhanced monomolecular lifetime for GBP films without a C 60 layer. Yet, despite the apparently improved bulk recombination properties, GBP PSCs remain severely limited by interfacial recombination at the perovskite/C 60 interface, as exemplified by the lower device V OC and FF. Interestingly, n id of complete PSC layer stacks shows a very similar trend, with Ref PSCs exhibiting n id of 1.81, which considerably reduces to 1. 53 (Fig. 3d). Critically, GBP&SP PSCs again show by far the lowest n id of 1.37, only slightly higher as for the half-layer stack without C 60 . This is among the lowest n id values reported in the literature for p-i-n PSCs with a PCE 4 20% and slightly below the value reported by Al Ashouri et al. for 2PACz-based wide-bandgap PSCs. 26,31,34,38 We note that when increasing the PEACl concentration to 3 mg ml À1 , n id of SP and GBP&SP PSCs increases again to 1.49 and 1.42 (Fig. S17, ESI †), respectively, showing that while the V OC of the respective PSCs further increases (Fig. S4, ESI †), the recombination behaviour does not further benefit from too thick passivation layers, to some extent contributing to the reduced FF. Finally, using the ideality factor and PLQY values we can determine the implied PCE and FF without and with C 60 and compare it with an ideal device (i.e., n id = 1) that has the same bandgap and J SC . 34 This allows us to estimate the remaining losses in our devices, i.e. the FF losses due to series/shunt resistance, non-ideal n id and non-radiative recombination. 34,125 As shown in Fig. S18  . This relates to a FF loss due to series/shunt resistance of roughly 4.2% (3.4%), while non-ideal n id and non-radiative recombination account for another 2% (1.1%) from C 60 and 5.5% (2.6%) from the bulk. This analysis reveals that, in addition to strongly enhancing the V OC , our dual passivation approach (GBP&SP) reduces the FF losses by 4.6% absolute as compared to Ref PSCs due to a reduced series/shunt resistance (0.8%) as well as simultaneous passivation of the perovskite/C 60 interface (0.9%) and grain boundaries (2.9%).
In summary, the TRPL and PLQY results show that (i) the perovskite/C 60 interface limits the V OC of our PSCs and (ii) that grain boundary passivation becomes specifically crucial in the case where the perovskite/CTL interfaces are already well passivated. Therefore, in line with the device data presented, we find that dual passivation is required to reach both the lowest non-radiative recombination losses and the lowest n id that only together result in the highest V OC and FF.

Material characterization
Having demonstrated that the superior performance of our dual passivation strategy stems from reduced non-radiative recombination at the grain boundaries as well as at the perovskite/C 60 interface, the question arises: How do SP and GBP modify the perovskite film morphology, structure or composition? To start with, we examine the films by SEM and identify a similar surface morphology and grain size distribution for the Ref and SP perovskite films (see Fig. S19a and b, ESI †). This observation is in line with the literature as surface treatment with low concentration long chain alkylammonium salts commonly does not alter the perovskite film morphology; more distinct changes to the morphology are only observed for higher concentrations (see Fig. S20, ESI †). 59,60,62,64,65,90 Further analysis of atomic force microscopy (AFM) images reveals a slight reduction in the root-mean-square surface roughness, which we attribute to the fact that PEACl preferentially fills regions close to the grain boundaries (Fig. S21, ESI †). 10,14 For all perovskite films, the grains that appear brighter in SEM are attributed to PbI 2 -rich crystallites as will be discussed later in more detail. 17,95 By incorporating PEACl:PbCl 2 in the film for GBP films, the size of the perovskite grains remains largely unchanged, however, the number and size of the PbI 2 -rich grains slightly increase (Fig. S19c, ESI †). Interestingly, notable small bright crystallites appear on the surface of the perovskite film which are specifically embedded close to the grain boundaries. This indicates that PEACl:PbCl 2 leads to passivation mainly near the grain boundaries. Upon additional surface treatment with PEACl for GBP&SP films, the size of these small bright crystallites is reduced and they appear more dispersed all over the surface, growing with a plate-like appearance perpendicular to the perovskite grains (Fig. S19d, ESI †). This implies that some reaction with these crystallites occurs when PEACl is deposited on top of GBP films.
To gain a better understanding about the phase or composition of the small and large bright grains observed in SEM at the surface of the perovskite films as well as their potential relevance in the context of this work, we carry out CL measurement. 126 For the Ref film, the grains which appear darker in SEM (highlighted by a green circle in Fig. 4) exhibit higher CL intensities (Fig. 4b) with the emission peak located at B774 nm (Fig. S22, ESI †) that correlates with the 3D perovskite phase with a bandgap of B1.57 eV (note that the CL setup is not spectrally calibrated). The grains which appear brighter and exhibit a different texture in the SEM images (highlighted by a yellow circle in Fig. 4; see further top-view SEM images in Fig. S23, ESI †) demonstrate lower CL intensities (dark spots in Fig. 4b) with a CL emission peak around 500 nm (Fig. S22, ESI †). By applying a 500 nm AE 40 nm bandpass filter to record the CL image, these grains can be clearly distinguished from the 3D perovskite grains (Fig. 4c). These regions are therefore attributed to PbI 2 -rich crystallites, in line with previous reports, 17,64 and as seen from cross-sectional SEM images, they appear to be located on top of 3D perovskite grains (Fig. S24, ESI †). Looking specifically at the CL signal from individual large PbI 2 -rich grains, there is indeed still a signal from the (underlying) 3D perovskite phase (Fig. S22, ESI †). For the SP film, we observe a slight charging of the SEM images (Fig. 4d), which we attribute to the insulating nature of a 2D RP phase forming at the surface. No noticeable change in the CL images of the SP film without and with 500 nm AE 40 nm bandpass filter is observed compared to the Ref film (Fig. 4e and f). This indicates that the large PbI 2 -rich grains are not completely chemically reacting upon PEACl surface treatment, which is in contrast to previous observations that bright PbI 2 related grains vanish upon treatment with various organic halides. 11,20,66,77,78,84 Nevertheless, due to the passivation effect, the CL signal of the 3D perovskite phase exhibits a much higher intensity compared to the Ref film (Fig. S25, ESI †), in line with the PLQY results. For the GBP film, the grains which appear as small bright grains close to the grain boundaries in SEM (highlighted by a red circle in Fig. 4g) are detected as dark small spots around the perovskite grains in the CL image (Fig. 4h). This stresses that these small grains exhibit lower CL intensities similar to the large PbI 2 -rich grains. Interestingly, when we apply the 500 nm AE 40 nm bandpass filter, only the features related to PbI 2 -rich grains can be observed, whereas no signal or feature correlated with the small bright grains is traceable (Fig. 4i). This indicates that these are not related to PbI 2 or PbCl 2 . Finally, for GBP&SP film no small grains are visible in SEM anymore due to their dispersion after surface treatment (Fig. 4j) and thus can no longer be identified in the CL image either without or with the bandpass filter ( Fig. 4k and l). Consistent with the PLQY and TRPL results, the GBP&SP film (Fig. 4k) exhibits the highest CL intensity at a wavelength of B774 nm (Fig. S25, ESI †).
This enhancement is attributed to the passivation of various recombination centres in the grain boundaries and/or at the surface of the perovskite layer. 126 Importantly, an additional CL peak at B620 nm appears, which cannot be related to either the 3D perovskite phase or the PbI 2 -rich phase. We find similar peaks in PL measurements of GBP&SP films as well as for SP films when using a higher PEACl concentration of 3 mg ml À1 (see Fig. S26, ESI †). We correlate this observation to the formation of a thin emissive 2D (PEA) 2 (Cs y FA 1Ày ) nÀ1 Pb n (I 1Àx Cl x ) 3n+1 RP phase with n = 2 at the surface of the films, as will be discussed in the following. 75,127

View Article Online
To analyse the crystal structure of our films, we perform X-ray diffraction (XRD) measurements (Fig. S27a, ESI †). All perovskite films exhibit the expected peaks at 14.21, 20.11, 24.61 and 28.41 from the (100), (110), (111) and (200) crystal planes of the 3D cubic a-Cs 0.18 FA 0.82 PbI 3 phase, 25,112,128 as well as a peak at B12.91 related to PbI 2 . The peak positions, intensities and FWHM of the XRD peaks are largely unchanged for SP films as observed in our previous work when using BABr for surface treatment. 59 For GBP films the intensity of the PbI 2 peak slightly increases (Fig. S27a, ESI †), in agreement with the larger number of PbI 2 -rich crystallites observed in SEM and CL. Furthermore, the ratio of the (100) to (111) peak slightly decreases for GBP films as compared to Ref and SP films, implying a slightly less preferred (100) orientation of the perovskite grains (Fig. S27b, ESI †). 25,99 Similar to SP films, GBP&SP films exhibit a largely unchanged XRD spectrum as compared to GBP films. We do not observe a signal related to a 2D RP phase for the SP and GBP&SP films, which could be related to either the passivation layer being too thin to be detected by XRD, the presence of a heterogeneous distribution of (PEA) 2 (Cs y FA 1Ày ) nÀ1 Pb n (I 1Àx Cl x ) 3n+1 phases with various n, or the presence of a non-crystalline PEA-based passivation layer. 14,59,68,69 Yet, upon further increasing the PEACl concentration to 4.5 mg ml À1 or 10 mg ml À1 , peaks at B5.31, 10.51, 15.71, 20.91, 26.11 and 31.81 start to appear. These can be attributed to a pure 2D (n = 1) (PEA) 2 Pb(I 1Àx Cl x ) 4 RP phase with a superlattice spacing of B1.7 nm that forms at the surface of the perovskite films (Fig. S28, ESI †). 77,[129][130][131][132] We perform XPS measurements of ITO/2PACz/perovskite stacks to get a better understanding of the elemental composition at the surface of our perovskite films and prove the presence of a PEACl-based passivation layer. For the Ref film, the XPS core-level spectra in Fig. 5a-c and Fig. S29a-c (ESI †) for the different elements in the 3D Cs 0.18 FA 0.82 PbI 3 absorber show the expected peaks with binding energies of B138.7 eV (Pb 4f 7/2 ), B400.8 eV (N 1s), B288.6 eV (C 1s from FA's N-CQN bonding), B619.6 eV (I 3d 5/2 ) and B725.3 eV (Cs 3d 5/2 ). 133 The additional C 1s peak at B284.8 eV is attributed to adventitious carbon (sp 3 C-C bonding) at the film surface. 17,95 There is no signal related to a Cl 2p doublet at a binding energy of B198.6 eV and B200.2 eV despite using CsCl in the precursor solution which can be explained by the sublimation of FACl during the annealing process. 16,112,134 We note that we do not observe a peak related to metallic lead (Pb 0 ) at B137 eV which emphasizes the high quality of our reference perovskite films. 19,20,74,87 Upon PEACl:PbCl 2 incorporation and/or PEACl surface treatment (GBP, SP and GBP&SP films) there are three key observations as compared to the Ref film. Firstly, for BP films the Pb 4f 7/2 core level slightly shifts toward higher binding energies and exhibits an increased FWHM, while for SP and GBP&SP films the signal intensity at B138.7 eV decreases and the peak becomes broader and asymmetric toward the highenergy side. This asymmetry is stronger for GBP&SP films. Secondly, a Cl signal at B198.6 eV (Cl 2p 3/2 ) and B200.2 eV (Cl 2p 1/2 ) appears for all films, which is by far strongest for GBP&SP films. Thirdly, two new peaks at B285.2 eV and 286.7 eV appear which are similar in intensity for SP and GBP&SP films and only very weak for GBP films. These peaks can be related to the C 1s emission from PEA (C-C and C-N bonds, respectively) with the expected stoichiometric ratio of 7 : 1.
Comparing SP and GBP&SP films using a higher PEACl concentration of 3 mg ml À1 , we find even more pronounced changes in the Pb 4f core levels (Fig. S30, ESI †). We stress that these are not accompanied by changes in the peak position or shape of the I 3d 5/2 , Cs 3d 5/2 and C 1s core levels, which excludes the possibility of a shift due to electronic doping of the perovskite bulk. 87,95,135,136 We therefore relate the damping of the signal at B138.7 eV for SP and GBP&SP, together with the appearance of a second Pb component at B139.1 eV as well as a Cl 2p and PEA signal, to the formation of a thin PEACl-based passivation layer on the surface of the films that has a different chemical environment. The fact that for GBP the damping of the signal at B138.7 eV is less pronounced and the peak becomes broader fits with our observation from SEM and CL that passivation happens mainly close to the grain boundary regions. To further test this interpretation, we show XPS measurements of SP and GBP&SP films with a much thicker passivation layer (10 mg ml À1 ) in Fig. S31 (ESI †). A single Pb 4f 7/2 peak at B139.4 eV can be observed with no remaining signal at B138.7 eV. In addition, no signals related to FA and Cs are observed anymore, which is an indication for the formation of a 2D RP phase with n = 1 and the composition (PEA) 2 Pb(I 1Àx Cl x ) 4 , at the film surface that completely damps the Pb signal from the underlying 3D perovskite phase. This is in line with the appearance of the related XRD peaks discussed above. We note that the calculated atomic ratio of (I + Cl)/Pb is 4.71 (SP) and 5.14 (GBP&SP) and thus even larger than the expected 4, which could be related to excess PEAI or PEACl at the film surface possibly forming an amorphous phase as reported recently (see Table S4, ESI †). 14,68 Similar shifts of the Pb 4f binding energies have previously been observed when changing the halide, i.e. for MAPbCl 3 and PbCl 2 as compared to MAPbI 3 and PbI 2 , 137,138 possibly due to the higher electronegativity of Cl as compared to I.
The formation of an n = 1 2D RP phase has been proposed in several studies that employed PEAI or OAI for surface treatment. 14,61,65,67,70,77,79,80 However, for the thin passivation layer thicknesses studied in this work, the atomic ratio of Cs/Pb for SP and GBP&SP films first increases above the respective values for the Ref and GBP films (1.5 mg ml À1 ), and only starts to slightly decrease for a higher PEACl concentration of 3 mg ml À1 (Table S4 and Fig. S32, ESI †). This indicates that Cs is taking part in the formation of the thin 2D RP passivation layer as otherwise the signal should be strictly decreasing because of the damping overlayer. This strict decrease in intensity with the formation of a 2D passivation layer can be seen for the atomic ratio of FA/Pb. This points towards a 2D RP (PEA) 2 (Cs y FA 1Ày ) nÀ1 Pb n (I 1Àx Cl x ) 3n+1 phase with y -1 at the surface of our GBP, SP and GBP&SP films. 10,64,78,80,119,127,132 Another clear observation is that the intensity of the Cl 2p doublet is considerably higher for GBP&SP as compared to SP films (Fig. 5b), while the I 3d 5/2 peak shows the reversed trend (Fig. S29c, ESI †). Accordingly, we note larger atomic ratios of Cl/Pb for GBP&SP vs. SP while the atomic ratio of I/Pb is smaller (Table S4 and Fig. S32, ESI †). This suggests that the 2D RP (PEA) 2 (Cs y FA 1Ày ) nÀ1 Pb n (I 1Àx Cl x ) 3n+1 phase in case of GBP&SP films is more chloride-rich as compared to SP films (x -1). The hypothesis that an n 4 1 2D RP (PEA) 2 (Cs y FA 1Ày ) nÀ1 Pb n (I 1Àx Cl x ) 3n+1 phase exist for all thin passivation layers is further corroborated by our observation of a CL and PL peak at B620 nm for the SP (3 mg ml À1 ) and GBP&SP films (see Fig. S25 and S26, ESI †) which points toward the existence of n of presumably 2. 75,127 To proof the existence of n 4 1, we performed additional XRD measurements using a more sensitive setup. We specifically analyzed the low-angle region of the XRD spectrum (2y o 121) of GBP&SP films employing various concentrations of PEACL for surface treatment (3, 3.5, 4 and 4.5 mg ml À1 ) and used the (100) peak of a GBP film as reference point. As shown in Fig. S33 (ESI †), for all GBP&SP films we observe a clear peak at B5.11 and a corresponding one at B10.21 that correlate to an n = 1 2D RP phase, as well a small peak at B3.81 that correlates to n = 2. Upon increasing the PEACL concentration, the intensity of the n = 1 peak monotonically increases, while that for n = 2 is comparable in intensity for the lower concentrations and only slightly in-creases for 4.5 mg ml À1 . Based on the XPS and XRD results together with the existence of a PL signal at B620 nm as discussed above, we conclude that for thin passivation layers the surface consists of a mixture of n = 1 and n = 2 2D RP phases, while for thicker passivation layers n = 1 becomes dominant toward the film surface. In order to give a rough estimate for the film thickness and n of the passivation layers, we take the Pb 4f 7/2 peak which shows a clear indication for two phases and fit it with two components (see Fig. S34, ESI †): one for the 3D bulk phase (I) and one for the 2D RP surface phase (II). Here, for simplicity, we assume that the Pb 4f 7/2 binding energy related to the 2D RP phase is situated at 139.1 eV. This allows to separate the relative contribution of the 2D (II) and 3D (I) material to the XPS signal and to evaluate the signal damping of the 3D bulk phase caused by the 2D RP overlayer. From here, we estimate a thickness of the 2D RP phase for a PEACl concentration of 1.5/3 mg ml À1 of B0.7/2.4 nm for SP films and B1.5

View Article Online
GBP&SP films (see Table S5 and further information in the XPS/UPS section in the ESI †). Finally, for simplicity assuming that the complete PEA signal measured in XPS is bound in a 2D RP phase, we can make a rough estimation of the respective n (averaged over the measured XPS spot) by analysing the atomic ratio of PEA to the 2D Pb (II) signal for 1.5/3 mg ml À1 PEACl (see Table S5, ESI †). We find that n for GBP&SP films (B2.6/2.3) is larger than for SP films (B1.2/1.7).
To assess the effect of the thin (PEA) 2 (Cs y FA 1Ày ) nÀ1 Pb n (I 1Àx Cl x ) 3n+1 surface layer on the energetics of our films we perform UPS measurements. From the onsets of the secondary electron cut-off and valence band spectra (Fig. S35, ESI † and Fig. 5d) we derive the energy band diagrams shown in Fig. 5e. The Ref films exhibit a WF of 4.68 eV and an ionization potential (IP) of 5.71 eV which represents a slightly n-type perovskite film, in line with previous observations. 24,25,66,95,139 For GBP films, the WF considerably decreases to 4.39 eV while the IP stays roughly constant at 5.69 eV, implying that the perovskite becomes more n-type, which could be attributed to a reduced electron trap density, in line with our SCLC results, or a different surface termination. 24,25,66,106,140 The SP and GBP&SP films exhibit a further reduction in WF to 4.23 eV and 3.91 eV, respectively, together with a similar IP of 5.41 eV and 5.44 eV. These changes we attribute to the formation of the thin 2D RP phase with a larger bandgap and different chemical environments at the film surface as observed by CL and XPS. This is further supported by measurements of SP and GBP&SP films with a PEACl concentration of 3 mg ml À1 , for which the WF further decreases to 3.89 eV and 3.82 eV, while the IP exhibits similar values of 5.45 eV and 5.47 eV, which implies a valence band onset of 1.57 eV and 1.66 eV, respectively, the latter being larger than the 3D perovskite bandgap (Fig. S36, ESI †). We also note that the shape of the VB density of states clearly is affected by surface treatment (Fig. 5e), showing that the electronic properties of the 2D RP surface layer are different compared to the 3D bulk perovskite. 141 Therefore, in addition to the expected chemical passivation, 53 we speculate that the 2D RP phase at the film surface with an increased distance of the VB to the Fermi level results in hole blocking and thus a reduced probability for holes in the 3D perovskite absorber to recombine with electrons in the C 60 layer. 25,53,72,86,142 In our case, electrons can effectively tunnel through the very thin (B0.7-1.5 nm) surface layer into C 60 , 25,38,142 resulting in still efficient charge extraction that allows very high FF for low PEACl concentrations of 1.5 mg ml À1 . For higher PEACl concentration of 3 mg ml À1 the passivation layers become too thick (B2.4-4.3 nm), making tunneling less likely which results in a decreased device performance due to a lower FF (Fig. S4, ESI †). Finally, we note that our results indicate band bending at the narrow 2D/3D interface as has recently been shown experimentally for the n-i-p architecture, 139,143 which also could contribute to the enhanced device performance. 10 So far, our analyses clearly show that our dual passivation strategy effectively reduces non-radiative recombination mainly at the perovskite/C 60 interface, but also the grain boundaries. Moreover, it is evident that a very thin 2D RP interlayer with a lower WF forms on the surface. The remaining questions are whether the PEACl:PbCl 2 additive passivates mainly defects in the grain interior and/or at the grain boundaries and how heterogeneous the surface passivation is for the different strategies. To shed light on this, we did frequency modulated Kelvin probe force microscopy (FM-KPFM) in the heterodyne KPFM implementation. KPFM measures the local contact potential difference (CPD) between a metallic tip and the sample surface and thus is directly related to the WF. 144,145 We used KPFM to map the effects of passivation on the CPD distribution of ITO/2PACz/perovskite/C 60 layer stacks, especially looking at grain boundaries, PbI 2 -rich grains and extent of heterogeneity (see Fig. 6). We want to stress that KPFM is prone to crosstalk from topography, often leading to KPFM contrast in strongly curved surface regions, such as grain boundaries. To minimize crosstalk artefacts, we use FM-KPFM 146 and carefully analysed the images, comparing the KPFM signal with the topography at the grain boundaries.
The map for the Ref sample (Fig. 6a) shows perovskite grains with a rather uniform CPD of (À430 AE 40) mV interrupted by grains with a less uniform and B110-230 mV lower CPD. The size and the surface distribution of the darker regions correspond to the PbI 2 -rich grains as observed in SEM and CL ( Fig. 4 and Fig. S19, ESI †). Here, the negative CPD contrast could be explained by a higher WF in PbI 2 as compared to the 3D perovskite due to its larger bandgap and p-type characteristics, 137,[147][148][149] in line with previous observations. 150,151 The map of the GBP sample (Fig. 6b) shows a similar trend in CPD contrast between more homogeneous perovskite grains with a CPD of (À700 AE 120) mV and less uniform spots with a CPD of (À1300 AE 280) mV. We note that the absolute value of the CPD depends on the tip's WF, which is sensitive to tip wear and contamination during the scanning. This could explain the overall lower absolute CPD in the GBP films. Assuming that the PbI 2 covered regions for the Ref and GBP sample have a comparable WF, we can use these ''dark'' regions as internal reference surface. Therefore, we first compare the relative change of the CPD value at the 3D perovskite grains with that at the dark PbI 2 -rich grains for the Ref as compared to the GBP samples, i.e. DCPD 3D-PbI2 (CPD 3D-grains -CPD PbI2-grains ). For the Ref sample DCPD 3D-PbI2 is B170 mV, while it is B600 mV for the GBP sample (note the different scales in Fig. 6a and b). This relative difference of DCPD 3D-PbI2 of B430 mV is comparable with the reduction in WF of B290 mV for GBP films as determined by UPS. Therefore, we attribute this observation to the fact that the PEACl:PbCl 2 additive mainly lowers the WF of the 3D perovskite grains, while the PbI 2 -rich grains are not modified by this strategy.
To analyse if the CPD at the grain boundaries with respect to the grain interior is modified for the GBP sample as compared to the Ref sample, i.e., DCPD GB (CPD grain boundary -CPD grain ), we show line profiles across representative grain boundaries in previously reported. [152][153][154][155] Such observations have been generally attributed to an enhanced ion and/or defect density, possibly due to a lower energy for defect formation at grain boundary regions. 42,46,154 In contrast, for the GBP sample we observed on average less contrast between grain boundaries and grain interior such that some of them cannot be clearly distinguished in the CPD map and line profiles anymore. We analysed in total a larger number of grain boundaries (B60) ( Fig. S38a and b, ESI †) for better statistics and found an average reduction from B93 mV to B74 mV for the GBP as compared to the Ref sample together with a large number of grain boundaries not showing any CPD contrast (see histogram in Fig. S38c, ESI †). A reduction in DCPD GB for the GBP as compared to the Ref sample indicates that the PEACl:PbCl 2 additive could specifically passivate the grain boundaries, resulting in a slightly more pronounced reduction of the WF with respect to the PbI 2 -rich grains as compared to the grain interior. We hypothesize that this is due to the formation of a PEACl-based 2D RP (PEA) 2 (Cs y FA 1Ày ) nÀ1 Pb n (I 1Àx Cl x ) 3n+1 phase specifically close to the grain boundaries, in line with the observations from CL and XPS. We further analyse the effects of PEACl surface treatment on the CPD (Fig. 6c and d). The SP Sample shows a heterogeneous CPD of (À52 AE 50) mV with no clear distinction between highand low-CPD grains anymore; however, we still observe a slight grain boundary contrast. We attribute this to the very thin 2D RP phase on top of the perovskite film -as identified by XPS -that changes the electrical properties of the film surface, including the regions of the PbI 2 -rich grains. We speculate that PEACl cannot completely penetrate and thus passivate the grain boundary regions, explaining why we still observe a CPD contrast. The enhanced heterogeneity of the CPD over different grains as compared to the Ref sample could be explained by the fact that the 2D RP (PEA) 2 (Cs y FA 1Ày ) nÀ1 Pb n (I 1Àx Cl x ) 3n+1 phase exhibits various n, i.e. a mixture of n = 1 and n = 2 phases with slightly different WF at different regions of the films. For the GBP&SP samples, we observe an even more heterogeneous CPD of (À670 AE 70) mV compared to all other samples ( Fig. 6d and Fig. S39, ESI †). Critically, no grain boundaries and PbI 2 -rich grains can be identified in the CPD map anymore. Taking the CL, XPS and KPFM results together, this indicates that employing both PEACl:PbCl 2 as additive and PEACl for surface treatment leads to a thin heterogeneous 2D RP (PEA) 2 (Cs y FA 1Ày ) nÀ1 Pb n (I 1Àx Cl x ) 3n+1 phase at the grain boundaries and surface of the perovskite film with n of B1-2. Finally, we note that the CPD value can be affected both by the facet orientation as well as the existence of RP phases with various n, which makes it difficult to disentangle these effects in detail. 146,153,[156][157][158]

Stability of passivated PSCs
Defect sites at grain boundaries and surfaces accelerate the degradation of perovskite thin films, since defects facilitate the migration of charged defects and mobile ions. 25,73,159,160 Having demonstrated that our dual passivation strategy reduces defects at the grain boundaries and the surface of the perovskite film, the question arises whether the concept also serves to increase the stability, i.e. slows down the degradation of the perovskite films. For this purpose, we compare the activation energy for ion   (Fig. 7b). Furthermore, we investigate the operational stability of PSCs for which the perovskite/ETL interface is passivated using a thin evaporated LiF layer, which is often employed in high-efficiency p-i-n PSCs. 31,35,38,88,89,93 In comparison with our GBP&SP PSCs (Fig. S42, ESI †) the p-i-n PSC with LiF passivation layer degrades much faster, reaching 80% of the initial PCE already after 7 h of constant illumination. Similar reports on the fast degradation of LiF containing p-i-n PSC can be found in literature. 31,93 Next to improved operational stability under constant AM1.5G illumination, the GBP&SP PSCs demonstrate improved thermal stability as compared to Ref PSCs by tracking the photovoltaic performance of the devices after aging under 85 1C heating in the dark over 1000 h (see Fig. 7c).
To further evaluate the stability with regard to moisture, we exposed unencapsulated Ref PSCs and GBP&SP PSCs to a relative humidity of B50% in ambient atmosphere and at room temperature for 1 day. The photographs and absorption data exhibit no changes for the GBP&SP film, while the Ref film is entirely decomposed to PbI 2 (Fig. S43a, ESI †). The increased contact angle of water droplets from 55.91 for the Ref film to 781 for the GBP&SP film (Fig. S43b, ESI †) confirms the better moisture resistance capability in the case of GBP&SP, which is attributed to the presence of hydrophobic PEA+ cations at the surface and grain boundaries of the perovskite films that acts as a hydrophobic barrier. 25,71,76 We note that the formation of shallow iodine interstitials upon passivation with iodide-based passivation molecules has recently been proposed to cause accelerated degradation of FAPbI 3 perovskites. 71 Our chloridebased dual passivation approach might potentially mitigate this issue. In summary, the presented fundamental assessment of stability for PSCs employing our dual passivation strategy highlights the importance of passivating defects at both the surface and the grain boundaries of perovskite films for achieving PCSs exhibiting both high efficiency and stability.

Conclusion
In summary, we demonstrate a dual passivation strategy for p-i-n PSCs that simultaneously passivates defects at the perovskite/C 60 interface as well as in the grain boundaries using the long chain alkylammonium salt phenethylammonium chloride. We achieve a substantial enhancement in charge carrier lifetime and quasi-Fermi level splitting compared to reference films as well as to films with either individual grain boundary or surface passivation. The best PSC with dual passivation achieves a significant absolute enhancement in both V OC (76 mV) and FF (4.6%) compared to the best reference device. As a result, a remarkable stabilized PCE of 22.3% is demonstrated, one of the highest reported for p-i-n PSCs. We attribute this improvement in performance to the formation of a heterogeneous 2D RP (PEA) 2 (Cs y FA 1Ày ) nÀ1 Pb n (I 1Àx Cl x ) 3n+1 phase with nB1-2 at the surface and grain boundaries of the films, which leads to (1) efficient chemical passivation of grain boundary and surface/interface defects and (2) additional hole blocking at the perovskite/C 60 interface. Finally, we demonstrate that the activation energy for ion migration is strongly increased upon dual passivation, which is reflected by an enhanced device stability under maximum power point (MPP) tracking and prolonged heat treatment. This work highlights the importance of defect management by employing a proper material both for grain boundary as well as surface passivation for achieving high-efficiency and stable inverted p-i-n PSCs. Thereby this work makes a relevant contribution to the advance of perovskite-based flexible and tandem photovoltaics.

Conflicts of interest
There are no conflicts to declare.