This is the accepted version of the article: Xie H., Wang Z., Chen Z., Pereyra C., Pols M., Galkowski K., Anaya M., Fu S., Jia X., Tang P., Kubicki D.J., Agarwalla A., Kim H.-S., Prochowicz D., Borrisé X., Bonn M., Bao C., Sun X., Zakeeruddin S.M., Emsley L., Arbiol J., Gao F., Fu F., Wang H.I., Tie. Decoupling the effects of defects on efficiency and stability through phosphonates in stable halide perovskite solar cells. Joule, (2021). 5. : 1246 - . 10.1016/j.joule.2021.04.003. Available at: https://dx.doi.org/10.1016/j.joule.2021.04.003 Manuscript Click here to view linked References Decoupling the effects of defects on efficiency and stability through phosphonates in stable halide perovskite solar cells Haibing Xie,1† Zaiwei Wang,2† Zehua Chen,3 Mike Pols,3 Krzysztof Gałkowski,4 Miguel Anaya,4 Shuai Fu,5 Xiaoyu Jia,5 Pengyi Tang,1 Dominik Józef Kubicki,6,7 Anand Agarwalla,2 Hui-Seon Kim,2 Daniel Prochowicz,6,8 Xavier Borrisé,9 Carlos Pereyra,1 Micha Bonn,5 Shaik Mohammed Zakeeruddin,6 Lyndon Emsley,7 Jordi Arbiol,1,10 Hai I. Wang,5 Klaas-Jan Tielrooij,1 Samuel D. Stranks,4 Shuxia Tao,3 Michael Grätzel,6 Anders Hagfeldt,2* Monica Lira-Cantu1,2* 1. Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC and the Barcelona Institute of Science and Technology (BIST). Building ICN2, Campus UAB E-08193, Bellaterra, Barcelona, Spain. 2. Laboratory of Photomolecular Science (LSPM), Institute of Chemical Sciences and Engineering, School of Basic Sciences, Ecole Polytechnique Fédérale de Lausanne, CH-1015 Lausanne, Switzerland. 3. Center for Computational Energy Research, Applied Physics, Eindhoven University of Technology, Eindhoven, the Netherlands. 4. Cavendish Laboratory, University of Cambridge, Cambridge, UK. 5. Max-Planck Institute for polymer Research, Ackermannweg 10, Mainz, Germany. 6. Laboratory for Photonics and Interfaces (LPI), Institute of Chemical Sciences and Engineering, School of Basic Sciences, Ecole Polytechnique Fédérale de Lausanne, CH-1015 Lausanne, Switzerland. 7. Laboratory of Magnetic Resonance (LRM), Institute of Chemical Sciences and Engineering, School of Basic Sciences, Ecole Polytechnique Fédérale de Lausanne, CH-1015 Lausanne, Switzerland. 8. Institute of Physical Chemistry, Polish Academy of Sciences, Kasprzaka 44/52, 01-224 Warsaw, Poland. 9. Institute of Microelectronics of Barcelona IMB-CNM, C/del Til·lers., Cjampus Universitat Autònoma de Barcelona (UAB) 08193, Cerdanyola del Vallès (Bellaterra), Barcelona, Spain. 10. ICREA, Pg. Lluís Companys 23, 08010 Barcelona, Catalonia, Spain. † These authors contributed equally to this work. * Corresponding authors: anders.hagfeldt@epfl.ch; monica.lira@icn2.cat 1 Abstract The understanding of defects is of paramount importance for the development of stable halide perovskite solar cells (PSCs). However, isolating their distinctive effects on the device efficiency and stability is currently a challenge. Here, we report that adding the organic molecule 3- phosphonopropionic acid (H3pp) to the halide perovskite precursor solution results in unchanged overall optoelectronic performance while having a tremendous impact on the stability of the halide perovskite by means of ion immobilization. As a result, we obtained PSCs with ~21 % efficiency and outstanding operational stability, retaining nearly 100 % of the initial efficiency after 1000 h at the maximum power point under simulated AM1.5 illumination. The strong interaction between the perovskite and the H3pp molecule through two types of hydrogen bonds (H…I and O…H) from the phosphonate group allows immobilization of ions that leads to remarkable device stability. This binding mode results in shallow point defect passivation that has a significant impact on the device stability but not on the nonradiative recombination and device efficiency. We expect that decoupling the effect of defects on the efficiency and stability will have important implications for the current understanding and advancement of operational PSCs. 2 Photovoltaic (PV) energy technologies are crucial to transform our society into a low-carbon economy, enabling the production of electricity from sunlight. Single-junction halide perovskite solar cells (PSCs) have already achieved a certified power conversion efficiency (PCE) above 25 %1,2. Despite their excellent PV performance, operational stability remains the main obstacle for the deployment of this technology. The PCE and operational stability in PSCs are thought to be strongly interrelated through the defect density at the surface, grain boundaries, interfaces and in the bulk of the halide perovskite (HP)2. Defects can either boost charge transport (generating extra free charge carriers by doping) or impede it by trapping or scattering the carriers3. Defects associated with deep trap states in HP (e.g. undercoordinated halides ions, undercoordinated Pb2+ ions) promote non-radiative recombination, leading against a high photovoltage, hampering the ultimate device efficiency4. Shallow defects, on the contrary, are considered benign in classical inorganic semiconductors and have little contribution to non-radiative recombination, not strongly influencing the efficiency of the device5. This high tolerance to shallow defects is what makes the perovskite family known as defect-tolerant5. However, the ionic-electronic characteristics of HPs, together with their “soft material” nature, indicates a strong ionic component where charged shallow level traps with low formation energies (e.g. I or MA vacancies), results in unfavourable for the solar cell, causing ion migration, phase segragation and hysteresis4-6. This is especially detrimental for the PSC under non-equilibrium conditions, such as high temperature, electrical bias, continuous light irradiation or their combination, where ion migration dictates the long-term stability of operational PSCs3,7-9. The introduction of additives (organic or inorganic compounds) into the HP layer is a method that has been effectively applied to passivate defects to increase the efficiency of the solar cell10-12. The success of such an approach is not straight forward and depends on multiple factors like HP thin film preparation, nature of precursors, type of additives13-17. Although efficient PSCs can be fabricated following this strategy and the stability is enhanced if compared with the control device without additive, the real outcome is not always stable PSC as defined as those able to display less than 10 % degradation after 1000 h of continuous illumination under 1 sun18. Thus, in order to attain highly efficient and stable PSCs it is not only important to passivate defects responsible for deep level traps that results in high efficiency, It is also important to ensure ion immobilization through the passivation of defect-induced shallow level traps5,6. The ionic motion is defined by the energy barriers to ionic transport in the HP, where halide ions (I-) and organic cations (MA, FA) are the main mobile ions7,14,19,20. Different passivation and grain-growth methods have been developed to decrease the defect density and suppress ionic migration20. The binding mode between the additive and the HP plays a critical role in the optoelectronic properties 3 of the device and its long-term stability. Additives with functional groups S=O, P=O or C=O can control the HP crystallization, passivate uncoordinated Pb2+ defects or both. Hydroxyl (-OH) and carboxyl (-COOH) groups also act as Lewis bases passivating uncoordinated lead defects by forming Pb-O bonds21. In particular, the hydrogen bonding energy between organic cations and PbI64- octahedra, as well as Pb-I bonds, weakens under continuous illumination. This effect facilitates the migration of both the organic cations and the halide anions in an HP, which would accelerate its decomposition in the absence of stabilizing additives20. The chemical shift trend for an additive varies considerably depending on the functional group participating in the bonding with the HP. In PSCs, the most employed additives encompass functional groups such as –COOH or −NH3+, while –PO(OH)2 is less employed22,23 17,24. The peculiar behaviour of the phosphonic acid derivative has been related to a stronger H-donating character25, which also plays an essential role in charge injection. Bonding through the phosphonic acid group has been reported to be more stable than bonding through carboxylic or amine groups, although the charge injection rates can be affected by the tetrahedral phosphorous centre and loss of conjugation26. For example, Brown et al. analysed the advantages of both –PO(OH)2 and –COOH anchoring agents in solar cells: –PO(OH)2 was employed to increase the surface binding stability, and carboxylic acids were employed to increase the electron injection efficiency27. This indicates that the chemical bonding between the HP and the additive also plays an important role on the regulation of the activation energy of ion migration, which determines the long-term stability of PSCs9. However, despite all these research efforts, the understanding of the binding modes in perovskite-additive interactions and the correlation between defect passivation, non-radiative recombination and device efficiency and stability remain poorly understood. In this work, we study the addition of the 3-phosphonopropionic acid (H3pp) organic molecule to an HP absorber and investigate the effect that two different anchoring groups (phosphonate –PO(OH)2 and carboxylate –COOH) have on the performance and stability of PSCs. In order to ensure an accurate comparison between the different characterization methods carried out to materials, thin films and complete devices, all our studies were carried out at room temperature. We fabricate complete solar cells with the configuration FTO/c-TiO2/m-TiO2/HP/Spiro-OMeTAD/Au based on the multication perovskite absorber with the nominal formula Rb0.05Cs0.05MA0.15FA0.75Pb1.05(I0.95Br0.05)3 (RbCsMAFA)28 with and without the H3pp molecule. The PSCs showed excellent and similar overall optoelectronic performance, leading to PSCs with an open-circuit voltage of Voc ~1.1 V and a PCE of ~21 %. Surprisingly, stability analyses of the PSC with H3pp showed negligible loss of the initial efficiency after 1000 h under 1 sun and maximum power point (MPP) tracking, while the reference device (without H3pp) showed > 20 % loss under the same conditions. To understand the effect of the H3pp molecule on the optoelectronic properties of the perovskite and its stability, we employed 4 contactless high-throughput characterizations, including time-resolved photoluminescence (TRPL) spectroscopy, photoluminescence quantum efficiency (PLQE) measurement and optical-pump THz- probe (OPTP) spectroscopy. First-principles density functional theory (DFT) calculations, supported by Fourier-transform infrared spectroscopy (FTIR) analyses, demonstrate the presence of two types of hydrogen bonds (H…I and O…H) between the –PO(OH)2 functional group and the perovskite. These strong binding modes occur through shallow point defects (e.g., FA and I vacancies)29,30 that block or immobilize ions, resulting in remarkable device stability. This interaction is also benign to the electronic structure of the perovskite and is not related to nonradiative recombination, as opposed to other binding modes that influence electronic interactions and passivate by binding directly to vacancy/defect sites31-33. We expect our findings to enable decoupling of the effects of defects on the efficiency and stability to advance on the understanding of shallow and deep defect passivation, non-radiative recombination and PSC performance. Photovoltaic performance Different organic additives sharing the phosphonic, –PO(OH)2, and carboxylic acid, –COOH, functional groups with different structures and chain lengths (Supplementary Fig. 1) were initially selected. These two functional groups were chosen due to the known interaction with the HP and the transport layers26,27,34-37. In all cases, the PCEs were below those obtained for the reference RbCsMAFA device (Supplementary Fig. 1). The best PV efficiency was observed for H3pp, one of the smallest and simplest molecules in the catalogue we explored. After careful optimization of its concentration within the HP, the 1:500 H3pp:HP ratio resulted in PSCs with the highest PCE, comparable to that of the reference PSCs (Supplementary Fig. 2). A scanning electron microscopy (SEM) cross-sectional image of the HP:H3pp PSC is depicted in Fig. 1a. The champion PSC fabricated with H3pp shows PCEs of 21.22 % and 20.45 % in reverse and forward scans, respectively (Fig. 1b and Supplementary Fig. 3), with a stabilized PCE of 21.1 % (inset of Supplementary Fig. 3). The Jsc from the IV curves agrees with the integrated Jsc obtained from the incident photon-to-electron conversion efficiency (IPCE) analysis (Supplementary Fig. 4). The optimization of the fabrication process statistically yielded average PCEs of 20.2 ± 0.7 % and 20.6 ± 0.4 % and an average V oc of 1.12 ±0.01 V for the reference and H3pp:HP PSCs, respectively (Fig. 1c and Fig. 1d). Both types of PSCs also demonstrate similar values in their other PV parameters (current density (Jsc), fill factor (FF) and hysteresis index (HI)) (Supplementary Fig. 5). To evaluate the recombination in our PSCs, we carried out light intensity-dependent Voc analysis in complete devices. Analyses were carried out in devices with the configuration FTO/c-TiO2/m- TiO2/HP/Spiro-OMeTAD/Au, with and without the H3pp additive in the HP layer (Supplementary Fig. 5 6a). The diode ideality factor (nid) can be extracted from the relation Voc = Eg + (nidkT/q) ln(I/I0), where Eg is the band gap, k is the Boltzmann constant, T is the temperature, q is the elementary charge, and I0 is a reference intensity usually at one sun38. Ideality factors with nid =1.80 and 1.85 for PSCs with and without the H3pp were observed, suggesting very similar p-n junction quality and 39,40 trap-assisted recombination in both types of PSCs , in agreement with the similar Voc and efficiency observed. Transient photocurrent (TPC) and open-circuit voltage decay (OCVD) analyses also displayed similar carrier transport and recombination properties for the PSCs with and without the H3pp additive, consistent with the similar Jsc and Voc values obtained (Supplementary Fig. 6b-c). Figure 1. PV performance. a, SEM cross-sectional image of the PSCs with the H3pp:HP material as the absorber. b, Forward and reverse IV curves of the champion H3pp:HP PSC. c-d, PCE and Voc statistics of the reference (HP) PSCs (blue) and the H3pp:HP-based PSCs (orange). The boxes show the 25th and 75th percentiles, and the whiskers display the 5th and 95th percentiles. The median and mean are indicated by the dividing lines across the boxes and the open square symbols, respectively. The maximum and minimum values are 6 represented by the cross symbols. e, TRPL for the HP (blue) and the 1:500 H3pp:HP ratio (orange) on glass. f, OPTP for the HP (blue) and the 1:500 H3pp:HP ratio (orange) on glass. To understand whether the recombination behaviour of the photoexcited carriers unequivocally comes from the perovskite films, TRPL analyses were carried out on HP thin films (deposited on glass) with and without the H3pp molecule. Fig. 1e shows the TRPL decay of reference and optimized H3pp:HP (1:500 ratio) films measured at a low fluence (509 nm, 50 nJ/cm 2 per pulse, equivalent to ~1 sun). The monoexponential fits to the data result in characteristic lifetimes very similar for both samples of τ ≈ 230 ns41, 42. The charge carrier lifetime is strongly related to the defect/trap density of the material under study, especially the nonradiative recombination42. The PLQE at 1 sun reached values of 1.90 % and 1.94 % for the HP and HP:H3pp thin films, respectively, demonstrating the good quality of the films and corroborating the similar recombination process in both samples (Supplementary Table 1). TRPL measurements at higher light intensities (fluences) equivalent to 1, 5, 10 and 100 suns (Supplementary Fig. 7) indicate very similar responses for the samples with and without the H3pp molecule up to 5 suns (Supplementary Fig. 7a-b). The similar charge carrier lifetimes at low fluences indicate that the addition of the H3pp molecule does not reduce nonradiative recombination in the perovskite films. However, for higher light intensities, i.e., 10 and 100 suns, we observed longer lifetimes, which could indicate a slight reduction in nonradiative recombination for the sample treated with H3pp in comparison to the neat HP films (Supplementary Fig. 7c-d). The comparable nonradiative recombination observed for the thin films with and without H3pp addition is supported by the comparable PLQEs and estimated quasi-Fermi level splitting values (QFLS, upper limit for Voc) in the perovskite films and PSCs (Supplementary Table 1-a and b, respectively)42,43,44. To probe the carrier dynamics on a shorter time scale, we employed OPTP spectroscopy with subpicosecond temporal resolution. In this experiment, we probed light-induced changes in the conductivity – i.e., the photoconductivity – as a function of the pump-probe delay time45. We carried out measurements on the structure of silica glass/perovskite for both the reference HP and the HP:H3pp sample (Fig. 1f). We observed that within a few picoseconds after excitation with the pump pulse (15 μJ/cm2, 800 nm), the photoconductivity is positive, indicating the presence of light-induced mobile charges. We observe a second-order recombination process caused by nongeminate electron-hole pair recombination46, leading to a reduction in the photoconductivity that is observable within a nanosecond. This process is followed by relaxation on a timescale of many nanoseconds (as probed using TRPL), which manifests as a constant long-term background in the background conductivity in the short OPTP time window (see also Supplementary Fig. 8). We also 7 observe that the normalized dynamics are nearly identical for the samples regardless of the presence of the H3pp additive. In summary, our results show very similar values of Voc and PCE (Fig. 1 and Supplementary Fig. 5) for the PSCs regardless of the presence of H3pp, which is ascribed to the negligible impact of H3pp on the nonradiative recombination of the HP and devices at low fluences (1 sun). Slight differences in the optoelectronic properties (longer lifetimes) of the H3pp:HP thin film are observed only at higher fluences (Supplementary Fig. 7-8); however, these values are not convincing enough to firmly state that the addition of H3pp reduces nonradiative recombination. Given that the nonradiative recombination is closely related to deep level defects that act as recombination centres, we speculate that defects (especially deep defects) are not passivated or that their passivation is negligible in our thin films when the H3pp additive is employed. Device stability and ion immobilization We sought to evaluate any possible effect of H3pp on the operational stability and ion migration of the PSCs (Fig. 2a). For this purpose, we followed the ISOS protocols recently reported for PSCs under 1 sun continuous illumination for 1000 h in N2 (ISOS-L-1l)18. First, we carried out stability analysis on the reference PSCs employing as the absorber the RbCsMAFA (●), without additive (blue) and PSCs with an optimal H3pp:HP ratio of 1:500 (orange) (Fig. 2a). The PSC with the optimal 1:500 H3pp:HP ratio retained nearly 100 % of the initial PCE, while the stability observed for the reference PSC showed about 28 % performance loss after 1,000 h under the same operational conditions (Fig. 2a). Interestingly, we replicate the same trends in other perovskite compositions such as Cs0.05MA0.15FA0.80Pb1.05(I0.85Br0.15)3 (CsMAFA) (Fig. 2a, ■, and Supplementary Fig. 9). The main difference between PSCs with the two HP is the initial few hours of the stability analyses for the PSC with the CsMAFA, where a decrease in PCE is initially observed and stabilizes with time. This response has been observed before and is attributed to the charge accumulation at the interface with Spiro-OMeTAD33,47. Interesting, our results demonstrate that the addition of H3pp can be extended to different HP absorbers with similar long-term stability results. To further understand the outstanding stability of the PSCs when we apply the H3pp molecule, we performed OPTP, PLQE and TRPL measurements on the perovskite films with and without H3pp on glass over time at high light intensities (Fig. 2b and c). The use of a high light intensity has been proposed as a way to accelerate the degradation of PSCs and to predict the solar cell lifetime since high intensity illumination of several suns can tremendously accelerate the degradation of solar cell devices and materials18. Fig. 2b shows the OPTP measurements of the photoconductivity carried out 8 on the reference HP and H3pp:HP thin films under 4 suns for 10 h. Data were taken every 30 min. The results indicate that the sample with the H3pp additive exhibits a more stable and higher photoconductivity after 2 h of light irradiation compared with the reference sample, which shows a slow but steady decrease in the photoconductivity. The THz photoconductivity is proportional to the number of photoexcited electron-hole pairs and to the mobility of these charges. Thus it provides microscopic insights into the fundamental processes (charge creation and charge transport) of the solar cell. The long-term stability of this signal demonstrates that the same number of charges is created and their mobility stays constant. The monitored response of the steady-state photoluminescence (PL) peak intensity under continuous irradiation of 100 suns for ~10 min (500 sec) under a N2 atmosphere (Fig. 2c) also demonstrates the high stability of the HP with H3pp in comparison with the reference HP, which degrades almost immediately after only a few seconds. 9 Figure 2. Stability and ion immobilization. a, Operational stability of PSCs applying RbCsMAFA (●) and CsMAFA (■) with only HP (blue) and with the H3pp:HP (orange), under 1,000 h of continuous illumination at 1 sun under a N2 atmosphere using white LEDs and MPP tracking. The H3pp doping ratio is the optimal 1:500. 3 devices for each type in the CsMAFA case were employed. b, OPTP analyses of perovskite films on glass, carried out in N2 under 4 suns for 10 h. c, PL measurements carried out on perovskite films on glass in N2 under 100 suns for 500 sec. d, Frequency-dependent capacitance of PSCs under 1 sun. e, Temperature-dependent dark conductivity of HP thin films on glass. To analyse the role of ion migration in the samples, we applied electrochemical impedance spectroscopy (EIS) and temperature-dependent dark conductivity measurements to PSC samples and 10 HP thin films, respectively. Fig. 2d shows the frequency-dependent capacitance of the PSCs with and without the H3pp additive obtained from the Nyquist plot of the EIS spectrum. The capacitance in the low frequency range (1-100 Hz) is one order of magnitude lower for the device with the H3pp molecule, indicating an effective reduction in ion migration and charge accumulation at the interfaces28. Ion migration activation energy (Ea) analysis was performed on HP samples deposited on glass via temperature-dependent dark conductivity measurements (Fig. 2e) according to the Nernst– Einstein relation of σ(T) = σ0 /T* exp (-Ea/kT), where T is the temperature, k is the Boltzmann constant, σ0 is a constant, and the activation energy Ea can be extracted from the slope of the ln(σT) ～1/KT relation48. The Ea for the HP is 340 meV, while 430 meV is obtained for the sample with the H3pp molecule, indicating effective mitigation of ion migration. In summary, the addition of H3pp has no substantial beneficial effect on the efficiency for PSCs under equilibrium conditions (as-prepared). Notably, it brings remarkable long-term stability to the PSCs analysed under nonequilibrium operational conditions (continuous light irradiation). Analyses of the carrier recombination and photoconductivity under multi-sun illumination support the high stability observed with the use of the H3pp additive not only under 1 sun but also under severe light irradiation up to 100 suns. Our results also demonstrate that the good stability can be extended to different HP absorbers, in this case CsMAFA and RbCsMAFA. PLQE and OPTP analyses at low fluences (~1 sun) rule out any variation in the non-radiative defect concentration. However, analyses at increasing fluences (above 10 suns) result in a small decrease in the deep trap density, which has an enormous impact on the device stability but not on the device performance. Our results also demonstrate that ion immobilization is the mechanism behind the outstanding long-term stability of the PSCs when the H3pp additive is employed. Effect of H3pp on the perovskite thin film: microstructure and binding mode On the basis of the PV results and outstanding PV stability, we aim to understand the effect of the H3pp molecule on the morphology, structure and composition of the HP thin films. Fig. 3a shows the morphology of the HP thin films with and without the H3pp additive in top-view SEM images. The addition of H3pp did not promote grain growth in the HP thin films since very similar grain dimensions (~280 nm on average) are observed for both samples. This indicates that grain boundaries are not reduced upon the addition of H3pp, in agreement with the similar nonradiative recombination levels in both types of PSCs, as grain boundaries are reported to be the source of defects32,49. Fig. 3b shows a cross-sectional annular dark-field scanning transmission electron microscopy (ADF-STEM) image of the different layers of the PSC, where the selected area highlighted 11 by a blue rectangle is the absorber HP:H3pp layer with a grain boundary. Electron energy loss spectroscopy (EELS) was applied to obtain the chemical composition maps of the HP:H3pp layer. According to the EELS maps, phosphorus (from the –PO(OH)2 of H3pp) is present in the HP bulk and is slightly richer near the grain boundary, demonstrating that H3pp is relatively segregated. As will be discussed later, X-ray diffraction (XRD) and solid-state nuclear magnetic resonance (NMR) confirm that H3pp is not incorporated into the crystal lattice. The cross-sectional lamella may contain several grains in the electron beam direction, and when the surface of the grains is surrounded by H3pp (at the 3D level), the EELS maps will show phosphorus in the bulk and a slightly rich presence at grain boundaries. The XRD patterns of the RbCsMAFA perovskite thin films without H3pp and at the optimal H3pp:HP 1:500 ratio are shown in Fig. 3c. No new peaks or peak shifts are observed in any of the HP films with different H3pp concentrations (Supplementary Fig. 10), indicating that H3pp may not be incorporated into the perovskite crystal lattice. Further solid-state NMR studies confirm that H3pp is not incorporated into the crystal lattice and most likely resides in the grain boundaries and on the surfaces (Supplementary Fig. 11). The (-111) crystal plane for the H3pp-doped sample is enhanced in comparison with the reference sample (Fig. 3c). The impact of the crystal orientation on the PSC performance is currently under debate by the research community14,50, probably due to variations in the precursor composition and fabrication processes. In our case, a negligible influence is observed. Evidence of the interaction between H3pp and the HP is found with FTIR analysis (Fig. 3d). The peak at 1712 cm-1 corresponds to the antisymmetric C-N stretching (overlapping with the C=O stretching), and the peak at 1470 cm-1 corresponds to the symmetric NH3+ bending. After the introduction of the H3pp molecule into the HP layer, the stretching vibration of –P=O in H3pp shifts from 1268 cm-1 to 1243 cm-1, whereas that of –C=O remains almost unchanged (1710 cm-1 to 1708 cm-1), which indicates a strong interaction of H3pp with the HP through the phosphonate group and a weak interaction through the carboxylate group. 12 Figure 3. Impact of H3pp on the structure and composition of the HP thin film. Comparison of the microstructural analysis of the RbCsMAFA (blue) and H3pp:RbCsMAFA 1:500 ratio (orange) thin films: a, Top- view SEM image. b, ADF-STEM micrograph (lateral view) of the HP:H3pp 1:500 ratio PSC, and corresponding EELS chemical composition maps of the selected absorber region with a grain boundary (blue rectangle): Individual I (red), P (green) and I + P. The L-edge (132 eV) was applied for the P analysis in the elemental maps. c, XRD patterns. d, FTIR analysis of the HP (orange) and HP:H3pp (blue), indicating the peaks corresponding to the P=O stretching mode between 1220 and 1300 nm and the C=O stretching mode between 1600 and 1800 nm. 13 To obtain more atomistic information regarding the interaction mode of H3pp with the HP, we further investigated their interaction by using DFT. We chose α-FAPbI3 as a simple perovskite model (FA is the dominant cation in the perovskite applied in this work). We began by computing the binding energy (Ebinding) of the H3pp molecule at FAI- and PbI2- terminated FAPbI3 surfaces as Ebinding = EH3pp + Esurface- Esurface-H3pp, where Esurface-H3pp, EH3pp and Esurface are the total energies of the H3pp-FAPbI3 system, H3pp molecule and FAPbI3 surface, respectively. We considered a number of possible interaction modes between the two, including the interaction of O-FA, H-I, and O-Pb, where the first and second atomic species come from H3pp and the HP, respectively. All interaction modes are energetically favourable, with sizeable binding energies in the range of 1.6 eV to 3.1 eV (Supplementary Fig. 12 and Supplementary Table 2). Notably, H3pp tends to bind much more strongly to the FAI-terminated surfaces than to the PbI2-terminated surfaces, with all configurations having binding energies as high as 2.9 eV to approximately 3.1 eV. Fig. 4a shows the most favourable configuration (binding energy of 3.13 eV), where H3pp interacts with the HP such that –PO(OH)2 points towards a FAI-terminated perovskite surface and forms a total of three hydrogen bonds of two different types. In the first type of hydrogen bond, the hydrogen atom from –P-OH interacts with I- (H-I), and in the second type, the oxygen atom from – P=O interacts with two hydrogen atoms from one NH2 moiety of FA+ (O-H). From a charge density difference analysis (Fig. 4b), we observe that both types of hydrogen bonds are present due to a significant depletion of electrons from the H atoms to the O and I atoms. Because the involved H atoms come from both H3pp and the HP, significant charge redistribution occurs on both sides, confirming strong interface interactions. Because of the newly created hydrogen bonds with the perovskite, we expect a slight weakening of the P=O bond itself in H3pp. Indeed, the wavenumber of the stretching mode of the –P=O bond, as calculated by DFT, shifts from 1246 cm-1 to 1171 cm-1 (Fig. 4c), which is a sign of weakening of the bond. Note that we also find another similar strong interaction mode of –PO(OH)2 with the FAI-terminated perovskite surface, a second favourable configuration with a binding energy of 2.99 eV (Supplementary Fig. 12b and Supplementary Table 2), where only two hydrogen bonds (instead of the three in Supplementary Fig. 12a and Fig. 4a) are formed. The calculated wavenumber of the stretching mode -P=O for this structure also shifts to a lower wavenumber of 1207 cm-1. While the wavenumber of the stretching mode of the -P=O bond shift to lower wavenumbers, that of the -C=O stretching mode remains almost unchanged (Fig. 4c), indicating the absence of a strong interaction with the HP. Our experimental FTIR spectrum shown in Fig. 3d confirms both trends. Therefore, combining the DFT calculations and FTIR data, we can establish that the dominant mode of interaction between the H3pp molecule and the perovskite is 14 through the formation of two types of hydrogen bonds between the –PO(OH)2 moiety and the FA+ cations and I- anions. Figure 4. Theoretical analysis of the interaction modes between H3pp and FAPbI 3. a, Example of one dominant interaction mode, where H3pp binds to the FAI-terminated surface of FAPbI3, forming three hydrogen bonds of two different types. b, Charge redistribution at the H3pp/FAPbI3 interface upon the adsorption of H3pp, where charge accumulation and depletion are indicated by yellow and cyan, respectively. c, Calculated FTIR wavenumbers of the stretching modes of the –C=O and –P=O bonds for H3pp and H3pp-modified FAPbI3. d, Formation energy of surface FA and I vacancies in H3pp-modified FAPbI3 compared with reference FAPbI3. Having established the dominant interaction modes of H3pp with the HP surface, we continue to study the effect of such interactions by calculating the formation energies of I anion and FA cation vacancies on the surface of the clean perovskite and on a surface that incorporates H3pp. As shown in Fig. 4d, the I and FA vacancy formation energies for a perovskite surface coordinated with H3pp species are increased by 2.03 and 1.20 eV, respectively, compared to the pristine perovskite. The second favourable configuration also gives slightly smaller (compared with the most favourable configuration) but still significantly large defect formation energy differences of 1.99 eV (for I vacancies) and 0.84 eV (for FA vacancies) compared with the pristine HP. The presence of H3pp 15 therefore inhibits the formation of I and FA vacancies at the surface. This result illustrates the mechanism behind ion immobilization and supports the excellent operational stability of the H3pp- modified PSC (Fig. 2a), where the creation of I and organic cation vacancies is typically believed to be responsible for the migration of ions in the perovskites under operational stress, such as electrical bias and light excitation51. In brief, the incorporation of H3pp within the HP layer leads to high nanostructure quality as a thin film absorber but does not translate into large grain sizes of the H3pp:HP thin film. The strong interaction between the HP and H3pp is corroborated by FTIR and NMR analyses. Our combined experiments and DFT analysis point to a strong interaction of the H3pp species with the perovskite surface, particularly via the phosphonate group. Such strong interactions include two types of hydrogen bonds (H-I and O-H). All these interactions potentially prevent the creation and diffusion of several point defects (e.g., FA and I vacancies) at the surfaces and grain boundaries of the perovskite film, enabling ion immobilization and highly stable PSCs under operational conditions. The current understanding of point defects in PSCs indicates that passivation of deep defects (e.g., Pb-I antisite defects)13 reduces the nonradiative losses and has a direct impact on the device performance (Voc and efficiency)3-5. In our work, the PLQE and OPTP results at low fluences (~1 sun) rule out any variation in the non-radiative defect concentration. However, analyses at increasing fluences (above 10 suns) indicate a small decrease in the deep trap density, which has an enormous impact on the device stability but not on the device performance. This result is unexpected because the application of additives in high-quality PSCs usually enhances both the efficiency and stability, and the role of defects in these properties is often linked. In our work, the strong binding modes of H3pp with the phosphonate functional group passivate defects through shallow point defects (e.g., FA and I vacancies)29,30 that block or immobilize ions (for example, MA, FA or I). This type of binding mode is benign to the electronic structure of the perovskite and does not relate to nonradiative recombination, as opposed to other binding modes that influence the electronic interaction and passivate by binding directly to vacancy/defect sites31-33. We expect that our results will permit to increase the understanding between defect passivation and PSC operational stability: It is not only significant to passivate defects responsible for deep level traps, it is also important to consider defect-induced shallow level traps. Finally, the relation of defect passivation and ion immobilization in these additives is not a simple task. The effect of multiple factors are usually taking part on the stability enhancement process. For example, intrinsic factors such as the concentration, size and structure of the additive (as shown in Supplementary Fig 1), as well as extrinsic factors such as high temperature, humidity or electrical bias shoud also be part of future studies. 16 Conclusions In summary, by applying the H3pp molecule to the perovskite precursor solution, we fabricated solar cells with excellent optoelectronic properties (Voc of ~1.1 V and PCE of ~21 %), which are not affected by the H3pp additive. Remarkably, the PSCs display outstanding operational stability, retaining nearly 100 % of the initial PCE after 1000 h under MPP operational conditions and light irradiation. The H3pp additive strongly bonds to the perovskite through the –PO(OH)2 functional group via two types of hydrogen bonds (H…I and O…H), passivating shallow point defects (e.g., FA and I vacancies). Since the passivation does not occur through deep defects (e.g., Pb-I antisite defects), the nonradiative recombination is not mitigated, and the device performance (Voc and efficiency) remains unaffected by the presence of the H3pp additive. Such binding modes facilitate immobilization of halide anions and organic cations, blocking ion migration and improving the device stability. Our findings permit decoupling of the effects of defects on the stability and performance and advance on the understanding of the relation between shallow and deep defect passivation, non-radiative recombination and PSC stability. A further understanding of the nature of the different binding modes of the functional groups of additives in HPs will likely be essential for the development of stable operational PSCs. Methods Chemicals. PbI2 (TCI, >98%), PbBr2 (TCI, >98%), FAI (Greatcell), MAI (Greatcell), MABr (Greatcell), CsI (ABCR, 99.998%), RbI (ABCR, 99.998%), dimethylformamide (DMF) ( ACROS organics, 99.8%, extra dry), dimethyl sulfoxide (DMSO) (ACROS organics, 99.8%, extra dry), 2,2′,7,7′-tetrakis(N,N-di-p-methoxyphenyl-amine)9,9′-spirobifluorene (spriro-OMeTAD) (Merck), 4-tert-butylpyridine (TBP) (Sigma Aldrich), lithium bis(trifluoromethanesulfonyl)imide (Li-TFSI) (Sigma Aldrich), titanium diisopropoxide bis(acetylacetonate) (Sigma Aldrich), acetylacetone (Sigma Aldrich), titania paste (Greatcell), chlorobenzene (Sigma Aldrich), 3-phosphonopropionic acid (Sigma Aldrich), acetonitrile (Sigma Aldrich). Device fabrication. FTO substrates (Nippon Sheet Glass 10 Ω/sq) were etched by Zn powder and 4M HCl, and then were cleaned with Hellmanex®, acetone and ethanol for 15 min, respectively inside an ultrasonic bath, finally were immediately dried with dry air. Compact TiO2 (c-TiO2) solution was prepared with titanium diisopropoxide bis(acetylacetonate): acetylacetone: ethanol = 0.6: 0.4: 9 (v : v) and then were sprayed onto FTO substrates at 450 °C.Mesoporous TiO 2(m-TiO2) paste was prepared with titanium paste: ethanol=1:6 (w:w), and was spin coated to c-TiO2 substrates at 5000 rpm for 20s and then annealed at 450 °C for 30 min. After that, the as-prepared FTO/c-TiO2/m-TiO2 substrates were quickly transferred to a N2 filled glove box for perovskite and spiro-OMeTAD deposition (no post-treatments were conducted for the substrates before deposition). The quadruple cation halide perovskite Rb0.05Cs0.05MA0.15FA0.75Pb1.05(I0.95Br0.05)3 was prepared as follows. Briefly, first 1.5 M stock solution of 1) CsI (DMSO), 2) RbI (DMSO), and 3) PbI2 (DMSO:DMF=1:4) were prepared, respectively. Then 1.5 M 4) (MABr)0.9(PbI2) (DMSO:DMF=1:4), and 5) (FAI)0.9(PbI2) (DMSO:DMF=1:4) were freshly prepared by dissolving MABr or FAI power in solution 3), respectively. After that, the solutions were mixed at a ratio of 4) : 5) : 1) : 2) : 3) = 190 : 950 : 60 : 60 : 60 (v:v) in sequence. The triple cation halide perovskite Cs 0.05MA0.15FA0.80Pb1.05(I0.85Br0.15)3 was prepared as follows. Briefly, first 1.5 M stock solution of 1) CsI (DMSO), 2) PbBr 2 (DMSO:DMF=1:4) and 3) PbI2 (DMSO:DMF=1:4) were prepared, respectively. Then 1.5 M 4) (MABr)(PbBr2) (DMSO:DMF=1:4) and 5) (FAI)(PbI2) (DMSO:DMF=1:4) were freshly prepared by dissolving MABr or FAI power in solution 2) and solution 3), respectively. After that, the solutions were mixed at a ratio of 4) : 5) : 1) : 3) = 150 : 800 : 50 : 100 (v:v) in sequence. 3-phosphonopropionic acid (H3pp) doped perovskite solutions were prepared by simply dissolving H3pp in the perovskite precursor solution to achieve H3pp: perovskite with various molar ratios (no heat or stirring is needed). The perovskite spin coating process was 17 carried out at 2000 rpm for 10s, and then 6000 rpm for 30s. Initially, 50 μL perovskite solution was dropped on a 1.5*2.5 cm2 FTO/c-TiO2/m-TiO2 substrate. During the second step of spin coating, 100 μL chlorobenzene was dropped at 15s before ending. The samples were annealed at 100 °C for 1 h on a hot plate for crystallization. Hole transporting layer was prepared by dissolving 0.12 g spiro-OMeTAD in 1130 μL chlorobenzene and then doped with 47.3 μL TBP and 23.5 μL Li-TFSI (1.8 M in acetonitrile). The spin coating was conducted at 4000 rpm for 20s with 50 μL solution. The finished devices were placed inside a dry air box for 12 hours to fully oxidize the spiro-OMeTAD. Finally, 80 nm Au was deposited as the front electrode by thermal evaporation. The evaporation rate was controlled in different stages to limit the damage to the spiro-OMeTAD layer: 0.003 nm/s for 0 - 1nm, 0.01 nm/s for 1- 10 nm, 0.02 nm/s for 10 - 20 nm and 0.06 nm/s for 20 - 80 nm. Solar cell characterization. The solar cells IV curves were measured using a 450 W Xenon light source (Oriel) equipped with an arc lamp housing (Newport, Model 66902, 50-500 W). The light intensity was calibrated with a Si photodiode equipped with an infrared cutoff filter (KG3, Schott), and was recorded during each measurement. I-V curves of the cells were obtained by a digital source meter (Keithley 2400). The voltage scan rate was 10 mV/s. The cells were masked to get 0.16 cm2 active area. Incident photon-to-electron conversion efficiency (IPCE) measurements were made based on a commercial apparatus (Arkeo-Ariadne from Cicci Research) using a 300 W Xenon lamp. All the measurements were conducted in air at room temperature. Solid-state NMR measurements. Starting materials were stored inside a glove box under argon. Perovskite powders were synthesized by grinding the reactants in an electric ball mill (Retsch Ball Mill MM-200 using a grinding jar (10 ml) and a ball (⌀10 mm) for 30 min at 25 Hz. The resulting powders were annealed at 140 °C for 10 minutes to reproduce the thin-film synthetic procedure. The amounts of reagents taken into the synthesis were as follows: α-FAPbI3: 0.172 g formamidinium hydroiodide (1.00 mmol) and 0.461 g PbI2 (1.00 mmol). H3pp-doped α-FAPbI3 (1:25 mol/mol): 0.172 g formamidinium hydroiodide (1.00 mmol), 0.461 g PbI 2 (1.00 mmol), and 0.006 g (0.04 mmol) H3pp. 14N (65.0 MHz) and 31P (364.4 MHz) NMR spectra were recorded on a Bruker Avance Neo 21.1 T spectrometer equipped with a 3.2 mm low-temperature CPMAS probe. 31P chemical shifts were referenced to 85% phosphoric acid. 14N spectra were referenced to solid NH4Cl (0 ppm) at 298 K. Low-temperature 1H-31P cross-polarization (CP) experiments used 1 ms optimized contact pulses, 4 s recycle delay, and 63 kHz SPINAL-64 1H decoupling. Echo-detected 31P spectra used the following optimized recycle delays: 1 s (α-FAPbI3 doped with H3pp), 300 s (neat H3pp). 14N spectra were acquired with recycle delays of 0.1 and 0.5 s, respectively. Fourier transform infrared spectroscopy (FTIR). FTIR was obtained by using TENSOR 27 FT-IR spectrometer (Bruker). Perovskite and H3pp-doped perovskite thin film were prepared following the same process as the devices, and then were scratched off carefully and mixed with KBr for measurements. To increase the signal of FTIR, concentrated H3pp doping was applied (H3pp: perovskite = 1:2.5). Electron energy loss spectroscopy (EELS). Focused ion beam (FIB, Zeiss Crossbeam 1560XB) was applied to prepare the lamellae for HRTEM measurements. 2 μm of Pt/carbon layer was deposited on top of the Au (active area) using electron beam evaporation to avoid the damage of the layer during FIB process. The initial cut for trenches was performed at 10 nA, then 2 nA and last polishing at 500 pA and 200 pA to limit the damage. Then the lamellae were lifted out and attached to the TEM grid. Final polishing was conducted in consecutive steps. The initial step was at 30 kV, 50 pA tilted at 57º (perpendicular to the FIB) and the final step was at 5 kV, 10 pA to avoid Ga + contamination as much as possible until the lamellae were transparent in the detector at 3 kV (electron acceleration voltage). The lamelIae are around 50 - 100 nm. To limit the possible degradation in air, fast transfer of lamellae in and out of the system was conducted (< 2 min). A grounded bracelet was used to take out the lamellae grid to avoid electrostatic damage. EELS analyses were obtained by using a FEI Tecnai F20 field emission gun microscope operated at 200 kV, which is equipped with a high-angle annular dark field (HAADF) detectorand embedded Quantum Gatan Image Filter. Images have been analyzed by means of GatanDigital Micrograph software. Scanning electron microscopy (SEM). SEM was performed using Quanta 650F with 5 kV accelerating voltage at a working distance of 8 cm. X-ray diffraction (XRD). X-ray diffraction (XRD) of the sample was measured using Empyrean (Panalytical) equipped with a ceramic tube (Cu anode, λ = 1.54060 Å) and a PIXel1D (Panalytical) detector. XRD data was analyzed using HighScore Plus software. Electrochemical Impedance Spectroscopy (EIS). Capacitance-Frequency measurements of perovskite solar cells were recorded with Paios (Fluxim, electrical measurement all-in-one platform for solar cells). The measurements are under a white LED source with calibrated light intensity close to one sun. Time-resolved photoluminescence (TRPL), Photoluminescence quantum efficiency (PLQE) and steady-state photoluminescence (PL)., PLQE and PL measurements were conducted by placing the samples in an integrating sphere. The different films were photoexcited by a 532-nm continuous-wave laser. The laser power was set at 60 mW/cm2 by using 18 a calibrated power meter, and it was monitored over time to ensure reproducible conditions for all the samples. Laser signal, laser signal attenuated by the sample (absorption) and photoluminescence signal were collected by an optical fibre coupled to an Andor iDus Si Detector, with an accumulation time of 30 s for each measurement. The optical response of the integrating sphere, the optical fibre and the detector were calibrated. The protocol followed for measuring and calculating the PLQE is based on reference52. For TRPL, a 509 nm pulsed (1 MHz repetition rate) semiconductor laser was used for excitation. The average laser power was set at 60 mW/cm2 (one sun equivalent intensity, or higher power equivalent to 5, 10 and 100 sun). A hybrid PMA photodiode with time resoloution of 25 ps was used as TCSPC detector. Samples were kept under N2 atmosphere until characterised to prevent them from degradation. Optical pump-THz probe (OPTP) spectroscopy. We perform photo-conductivity measurements of samples using optical pump-THz probe (OPTP) spectroscopy. In the setup, the THz generation is based on the optical rectification process in ZnTe non-linear crystals, using 800 nm pulses with duration of ~ 50 fs. The time-dependent electric field strength of THz pulses (~ 1-2 ps duration) transmitted the sample is detected by a second ZnTe crystal using electro-optic sampling. By fixing the sampling delay time to the peak position of the THz pulse, we monitor the THz absorption change by a third laser pulse with a wavelength of 800 nm. By tuning the time delay between the pump and THz probe, we measure the time- dependent THz absorption and therefore, the photo-conductivity of our sample in the time domain. The OPTP setup is kept in an N2 environment to avoid THz absorption in air. Temperature dependent dark conductivity. The conductivity of the perovskite with and without H3pp was measured at different temperatures by applying 2 V bias to glass/perovskite with electrode spacing of 10 μm on the surface, corresponding to an average electrical field of 0.2 V μm−1. Sample temperature was varied in the close-circle He cryostat and measured by Si-diode from 120 K to 330 K. We firstly cooled the device to 120 K and then stabilized it at 120 K for half an hour. Subsequently the device was heated to an objective temperature with an interval of 20 K and stabilized at the temperature for 10 minutes before the current measurement was performed. To avoid transient current spikes which occurred in the first few seconds after bias, the conductivity was extracted from the device current between 20 and 50 s after the bias was applied48,53-54. Light intensity dependent Voc, Transient photocurrent (TPC) and Open circuit voltage decay (OCVD). Light intensity dependent Voc, TPC and OCVD are analyzed at state-of-the-art devices (FTO/c-TiO2/m-TiO2/perovskite/spiro-OMeTAD/Au) using PAIOS (Fluxim, electrical measurement all-in-one platform for solar cells). The measurements are under a white LED source with varied light intensity from 0.3 to 1 sun for light intensity dependent Voc38 and 1 sun for TPC and OCVD. Stability Analyses. The stability measurements were carried out following ISOS-L protocol for PSCs18. The tests were performed under N2 atmosphere in a seald holder, tested under continuous light illumination and maximum power point tracking at room temperature (controlled and monitored to be 25 °C). The light source consisted of an array of white LEDs powered by a constant current. The LED type is LXM3-PW51 with an emission spectrum of 400-750 nm (see Supplementary Fig. 13). Equivalent sun intensities were calibrated using a calibrated Si reference diode equipped with a KG-5 filter. Computational methods. Since FA as a cation and I as an anion are the most abundant species in the experimentally studied (R0.05Cs0.05MA0.15FA0.75Pb1.05(I0.95Br0.05)3), the calculations were carried out using FAPbI3 as a model system. The surfaces were modelled using slab models consisting of (2×2) cells in the x and y directions and nine atomic layers of iodine constructed from the bulk structure with a vacuum of 15 Å in the z direction. Structural relaxation of all structures were performed within the framework of DFT, as implemented in the Vienna Ab Initio Simulation Package (VASP).55 The Perdew, Burke and Ernzerhof functional within the generalized gradient approach was used.56,57 During the structure optimization, five bottom atomic layers were fixed and ions of rest layers and H3pp molecule were allowed to relax. An energy cutoff of 500 eV and a k-point of 3× 3 × 1 achieve energy and force convergence of 0.01 meV and 20 meV Å-1, respectively. Information on chemical bonding analysis and formation energies of I - and FA+ vacancies in the reference, H3pp-containing perovskites is presented in Supplementary Information. Acknowledgements This article is based upon work from COST Action StableNextSol project MP1307, supported by COST (European Cooperation in Science and Technology). We thank the Spanish MINECO through the Severo Ochoa Centers of Excellence Program under Grant SEV-2013-0295 for the postdoctoral contract to H.X.. M.G. and S.M.Z. thank the King Abdulaziz City for Science and Technology (KACST) for the financial support. A. H. thank the financial support from the European Union H2020, ESPResSo project grant agreement 764047 and Swiss National Science Foundation project IZLCZ2_170177. Z.W. and S.F are thankful for the “China Scholarship Council” fellowship (CSC). H.-S.K., M.G. and S.M.Z. thank the financial support from the GRAPHENE Flagship Core 2 project supported by the European Commission H2020 Programme under contract 785219. P.T. and J.A. acknowledge funding from Generalitat de Catalunya 2017 SGR 327 and the Spanish MINECO project ANAPHASE (ENE2017-85087-C3). D.J.K. and L.E. acknowledge support from Swiss National Science Foundation Grant No. 200021_160112. S.T. acknowledges funding from the Computational Sciences for Energy Research tenure track programme of Shell, NWO and FOM (project no. 15CST04-2), the Netherlands. S.D.S. and M.A. acknowledge funding from the European 19 Research Council (ERC) (Grant Agreement No. 756962 [HYPERION]) and the Marie Skłodowska-Curie actions (Grant Agreement No. 841386) under the European Union’s Horizon 2020 research and innovation programme. S.D.S acknowledges support from the Royal Society and Tata Group (UF150033) and EPSRC (EP/R023980/1). K.G. appreciates support from the Polish Ministry of Science and Higher Education within the Mobilnosc Plus program (Grant no. 1603/MOB/V/2017/0). To the SolarEra.Net Cofund 2 (EU) and the AEI (Spain) for the project PrOperPhotoMiLe (Ref12). To the Spanish Ministry of Scence and Innovation for the grant GraPErOs (ENE2016-79282-C5-2-R), Self-Power (PID2019- 104272RB-C54) and the OrgEnergy Excelence Network (CTQ2016-81911-REDT). To the Agència de Gestiód'Ajuts Universitaris i de Recerca (AGAUR) for the support to the consolidated Catalonia research group 217 SGR 329 and the Xarxa de Referència en Materials Avançats per a l'Energia (Xarmae) and the Xarxa d’R+D+I Energy for Society (XRE4S). Part of this work is under Materials Science Ph.D. Degree for A.M., P.T. and C.P. of the Universitat Autonoma de Barcelona. We thank CONACYT for the scholarship to C.P. We acknowledge Libertad Sole and also the Clean-Room from IMB-CNM for FIB process. ICN2 is supported by the Severo Ochoa program from Spanish MINECO (Grant No. SEV-2017-0706) and is funded by the CERCA Programme / Generalitat de Catalunya. Authors contribution M.L.-C. conceived the idea.; M.L.-C. and A.H. supervised the study.; M.L.-C. and H.X. prepared the manuscript with support from all coauthors; H.X. and Z.W. designed the experiments, fabricated the devices and measured I-V data.; H.X. carried out different analyses including FTIR, XRD, SEM.; Z.C. and M.P. performed DFT calculations under supervision of S.T.; P.T. and J.A. performed EELS and did the relevant analyses.; D.J.K. and L.E. conducted the solid-state NMR and carried out the relevant analyses.; D.P. assisted with NMR samples preparation.; A.A. performed stability test under operational conditions; X.B. conducted the FIB process for lamellae.; C.P. helped with device optimization.; H.-S.K. assisted with device fabrication and data analysis. K.G., M.A. and S.S. carried out the TRPL,PLQE and PL analyses.; S.F, X.J., M.B., H.W. and K-J.T carried out the OPTP analyses. M.G. and S.M.Z. coordinated the work. Competing Interests The authors declare no competing financial interests. References 1 Green, M. et al. Solar cell efficiency tables (version 57). Progress in Photovoltaics: Research and Applications 29, 3-15, doi:https://doi.org/10.1002/pip.3371 (2021). 2 Kumar, S., Hodes, G. & Cahen, D. Defects in halide perovskites: The lattice as a boojum? MRS Bulletin 45, 478-484, doi:10.1557/mrs.2020.146 (2020). 3 Tong, C.-J., Li, L., Liu, L.-M. & Prezhdo, O. V. Synergy between Ion Migration and Charge Carrier Recombination in Metal-Halide Perovskites. Journal of the American Chemical Society 142, 3060-3068, doi:10.1021/jacs.9b12391 (2020). 4 Fu, L. et al. Defect passivation strategies in perovskites for an enhanced photovoltaic performance. Energy & Environmental Science 13, 4017-4056, doi:10.1039/D0EE01767A (2020). 5 Chen, B., Rudd, P. N., Yang, S., Yuan, Y. & Huang, J. Imperfections and their passivation in halide perovskite solar cells. Chemical Society Reviews 48, 3842-3867, doi:10.1039/C8CS00853A (2019). 6 Tan, S. et al. Shallow Iodine Defects Accelerate the Degradation of α-Phase Formamidinium Perovskite. Joule 4, 2426-2442, doi:https://doi.org/10.1016/j.joule.2020.08.016 (2020). 7 Zhang, T., Hu, C. & Yang, S. Ion Migration: A “Double-Edged Sword” for Halide-Perovskite- Based Electronic Devices. Small Methods 4, 1900552, doi:10.1002/smtd.201900552 (2020). 20 8 Tan, S. et al. Steric Impediment of Ion Migration Contributes to Improved Operational Stability of Perovskite Solar Cells. Advanced Materials 32, 1906995, doi:10.1002/adma.201906995 (2020). 9 Lee, J.-W., Kim, S.-G., Yang, J.-M., Yang, Y. & Park, N.-G. Verification and mitigation of ion migration in perovskite solar cells. APL Materials 7, 041111, doi:10.1063/1.5085643 (2019). 10 Zhang, H., Nazeeruddin, M. K. & Choy, W. C. H. Perovskite Photovoltaics: The Significant Role of Ligands in Film Formation, Passivation, and Stability. Advanced Materials 31, 1805702, doi:10.1002/adma.201805702 (2019). 11 Ran, C., Xu, J., Gao, W., Huang, C. & Dou, S. Defects in metal triiodide perovskite materials towards high-performance solar cells: origin, impact, characterization, and engineering. Chemical Society reviews 47 12, 4581-4610 (2018). 12 Rajagopal, A., Yao, K. & Jen, A. K.-Y. Toward Perovskite Solar Cell Commercialization: A Perspective and Research Roadmap Based on Interfacial Engineering. Advanced Materials 30, 1800455, doi:10.1002/adma.201800455 (2018). 13 Wang, R. et al. Constructive molecular configurations for surface-defect passivation of perovskite photovoltaics. Science 366, 1509-1513, doi:10.1126/science.aay9698 (2019). 14 Zheng, X. et al. Managing grains and interfaces via ligand anchoring enables 22.3%-efficiency inverted perovskite solar cells. Nature Energy 5, 131-140, doi:10.1038/s41560-019-0538-4 (2020). 15 Xie, H. & Lira-Cantú, M. Multi-component engineering to enable long-term operational stability of perovskite solar cells. J. PhysEnergy 2, 024008 (2020). 16 Pereyra, C., Xie, H. & Lira-Cantu, M. Additives for Stable Halide Perovskite Solar Cells. J. Energ. Chem. Accepted (2020.). 17 Braly, I. L. et al. Hybrid perovskite films approaching the radiative limit with over 90% photoluminescence quantum efficiency. Nature Photonics 12, 355-361, doi:10.1038/s41566-018- 0154-z (2018). 18 Khenkin, M. V. et al. Consensus statement for stability assessment and reporting for perovskite photovoltaics based on ISOS procedures. Nature Energy 5, 35-49, doi:10.1038/s41560- 019-0529-5 (2020). 19 Zhao, Y. et al. Mobile-Ion-Induced Degradation of Organic Hole-Selective Layers in Perovskite Solar Cells. The Journal of Physical Chemistry C 121, 14517-14523, doi:10.1021/acs.jpcc.7b04684 (2017). 20 Wei, D. et al. Ion-Migration Inhibition by the Cation–π Interaction in Perovskite Materials for Efficient and Stable Perovskite Solar Cells. Advanced Materials 30, 1707583, doi:10.1002/adma.201707583 (2018). 21 21 Choi, M.-J. et al. Functional additives for high-performance inverted planar perovskite solar cells with exceeding 20% efficiency: Selective complexation of organic cations in precursors. Nano Energy 71, 104639, doi:https://doi.org/10.1016/j.nanoen.2020.104639 (2020). 22 Blanchard, J. W., Groy, T. L., Yarger, J. L. & Holland, G. P. Investigating Hydrogen-Bonded Phosphonic Acids with Proton Ultrafast MAS NMR and DFT Calculations. The Journal of Physical Chemistry C 116, 18824-18830, doi:10.1021/jp305229s (2012). 23 Li, X. et al. Improved performance and stability of perovskite solar cells by crystal crosslinking with alkylphosphonic acid ω -ammonium chlorides. Nature Chemistry 7, 703-711, doi:10.1038/nchem.2324 (2015). 24 deQuilettes, D. W. et al. Photoluminescence Lifetimes Exceeding 8 μs and Quantum Yields Exceeding 30% in Hybrid Perovskite Thin Films by Ligand Passivation. ACS Energy Letters 1, 438-444, doi:10.1021/acsenergylett.6b00236 (2016). 25 Faucon, A. et al. Fluorescent carboxylic and phosphonic acids: comparative photophysics from solution to organic nanoparticles. Physical Chemistry Chemical Physics 15, 12748-12756, doi:10.1039/C3CP51369F (2013). 26 Guerrero, G., Alauzun, J. G., Granier, M., Laurencin, D. & Mutin, P. H. Phosphonate coupling molecules for the control of surface/interface properties and the synthesis of nanomaterials. Dalton Transactions 42, 12569-12585, doi:10.1039/C3DT51193F (2013). 27 Brown, D. G., Schauer, P. A., Borau-Garcia, J., Fancy, B. R. & Berlinguette, C. P. Stabilization of Ruthenium Sensitizers to TiO2 Surfaces through Cooperative Anchoring Groups. Journal of the American Chemical Society 135, 1692-1695, doi:10.1021/ja310965h (2013). 28 Kim, H.-S. et al. Control of I–V Hysteresis in CH3NH3PbI3 Perovskite Solar Cell. The Journal of Physical Chemistry Letters 6, 4633-4639, doi:10.1021/acs.jpclett.5b02273 (2015). 29 Ball, J. M. & Petrozza, A. Defects in perovskite-halides and their effects in solar cells. Nature Energy 1, 16149, doi:10.1038/nenergy.2016.149 (2016). 30 Yin, W.-J., Shi, T. & Yan, Y. Unique Properties of Halide Perovskites as Possible Origins of the Superior Solar Cell Performance. Advanced Materials 26, 4653-4658, doi:10.1002/adma.201306281 (2014). 31 Zuo, L. et al. Polymer-modified halide perovskite films for efficient and stable planar heterojunction solar cells. Science Advances 3, e1700106, doi:10.1126/sciadv.1700106 (2017). 32 Bi, D. et al. Multifunctional molecular modulators for perovskite solar cells with over 20% efficiency and high operational stability. Nature Communications 9, 4482, doi:10.1038/s41467-018- 06709-w (2018). 22 33 Wang, L. et al. A Eu3+/Eu2+ ion redox shuttle imparts operational durability to Pb-I perovskite solar cells. Science 363, 265-270, doi:10.1126/science.aau5701 (2019). 34 Luschtinetz, R., Gemming, S. & Seifert, G. Anchoring functional molecules on TiO2 surfaces: A comparison between the carboxylic and the phosphonic acid group. The European Physical Journal Plus 126, 98, doi:10.1140/epjp/i2011-11098-4 (2011). 35 Martini, L. A. et al. Modular Assembly of High-Potential Zinc Porphyrin Photosensitizers Attached to TiO2 with a Series of Anchoring Groups. The Journal of Physical Chemistry C 117, 14526- 14533, doi:10.1021/jp4053456 (2013). 36 Zhang, L. & Cole, J. M. Anchoring Groups for Dye-Sensitized Solar Cells. ACS Applied Materials & Interfaces 7, 3427-3455, doi:10.1021/am507334m (2015). 37 Bae, E. et al. Effects of Surface Anchoring Groups (Carboxylate vs Phosphonate) in Ruthenium-Complex-Sensitized TiO2 on Visible Light Reactivity in Aqueous Suspensions. The Journal of Physical Chemistry B 108, 14093-14101, doi:10.1021/jp047777p (2004). 38 Tress, W. et al. Interpretation and evolution of open-circuit voltage, recombination, ideality factor and subgap defect states during reversible light-soaking and irreversible degradation of perovskite solar cells. Energy & Environmental Science 11, 151-165, doi:10.1039/C7EE02415K (2018). 39 Yang, D. et al. High efficiency planar-type perovskite solar cells with negligible hysteresis using EDTA-complexed SnO2. Nature Communications 9, 3239, doi:10.1038/s41467-018-05760-x (2018). 40 Jiang, Q. et al. Surface passivation of perovskite film for efficient solar cells. Nature Photonics 13, 460-466, doi:10.1038/s41566-019-0398-2 (2019). 41 Stranks, S. D. et al. Electron-hole diffusion lengths exceeding 1 micrometer in an organometal trihalide perovskite absorber. Science 342, 341-344, doi:10.1126/science.1243982 (2013). 42 Stolterfoht, M. et al. The impact of energy alignment and interfacial recombination on the internal and external open-circuit voltage of perovskite solar cells. Energy & Environmental Science 12, 2778-2788, doi:10.1039/C9EE02020A (2019). 43 Kim, G. Y. et al. Large tunable photoeffect on ion conduction in halide perovskites and implications for photodecomposition. Nature Materials 17, 445-449, doi:10.1038/s41563-018-0038- 0 (2018). 44 Meggiolaro, D. et al. Iodine chemistry determines the defect tolerance of lead-halide perovskites. Energy & Environmental Science 11, 702-713, doi:10.1039/C8EE00124C (2018). 23 45 Becker, P. et al. Low Temperature Synthesis of Stable γ-CsPbI3 Perovskite Layers for Solar Cells Obtained by High Throughput Experimentation. Advanced Energy Materials 9, 1900555, doi:10.1002/aenm.201900555 (2019). 46 Manser, J. S. & Kamat, P. V. Band filling with free charge carriers in organometal halide perovskites. Nature Photonics 8, 737-743, doi:10.1038/nphoton.2014.171 (2014). 47 Liu, Z. et al. A holistic approach to interface stabilization for efficient perovskite solar modules with over 2,000-hour operational stability. Nature Energy 5, 596-604, doi:10.1038/s41560- 020-0653-2 (2020). 48 Yuan, Y. et al. Photovoltaic Switching Mechanism in Lateral Structure Hybrid Perovskite Solar Cells. Advanced Energy Materials 5, 1500615, doi:10.1002/aenm.201500615 (2015). 49 Shi, D. et al. Low trap-state density and long carrier diffusion in organolead trihalide perovskite single crystals. Science 347, 519-522, doi:10.1126/science.aaa2725 (2015). 50 Muscarella, L. A. et al. Crystal Orientation and Grain Size: Do They Determine Optoelectronic Properties of MAPbI3 Perovskite? The Journal of Physical Chemistry Letters 10, 6010-6018, doi:10.1021/acs.jpclett.9b02757 (2019). 51 Li, N. et al. Cation and anion immobilization through chemical bonding enhancement with fluorides for stable halide perovskite solar cells. Nature Energy 4, 408-415, doi:10.1038/s41560-019- 0382-6 (2019). 52 de Mello, J. C., Wittmann, H. F. & Friend, R. H. An improved experimental determination of external photoluminescence quantum efficiency. Advanced Materials 9, 230-232, doi:10.1002/adma.19970090308 (1997). 53 Xing, J. et al. Ultrafast ion migration in hybrid perovskite polycrystalline thin films under light and suppression in single crystals. Physical Chemistry Chemical Physics 18, 30484-30490, doi:10.1039/C6CP06496E (2016). 54 Zhao, J. et al. Strained hybrid perovskite thin films and their impact on the intrinsic stability of perovskite solar cells. Science Advances 3, eaao5616, doi:10.1126/sciadv.aao5616 (2017). 55 Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Physical Review B 54, 11169-11186, doi:10.1103/PhysRevB.54.11169 (1996). 56 Blöchl, P. E. Projector augmented-wave method. Physical Review B 50, 17953-17979, doi:10.1103/PhysRevB.50.17953 (1994). 57 Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys Rev Lett 77, 3865-3868, doi:10.1103/PhysRevLett.77.3865 (1996). 24 Supplemental Text and Figures Supplementary information Decoupling of defects from stability and performance through phosphonates in stable halide perovskite solar cells Haibing Xie,1 Zaiwei Wang,2 Zehua Chen,3 Mike Pols,3 Krzysztof Gałkowski,4 Miguel Anaya,4 Shuai Fu,5 Xiaoyu Jia,5 Pengyi Tang,1 Dominik Józef Kubicki,6,7 Anand Agarwalla,2 Hui-Seon Kim,2 Daniel Prochowicz,6,8 Xavier Borrisé,9 Carlos Pereyra,1 Micha Bonn,5 Shaik Mohammed Zakeeruddin,6 Lyndon Emsley,7 Jordi Arbiol,1,10 Hai Wang,5 Klaas-Jan Tielrooij,1 Samuel D. Stranks,4 Shuxia Tao,3 Michael Grätzel,6 Anders Hagfeldt,2* Monica Lira-Cantu1,2* 1. Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC and the Barcelona Institute of Science and Technology (BIST). Building ICN2, Campus UAB E-08193, Bellaterra, Barcelona, Spain. 2. Laboratory of Photomolecular Science (LSPM), Institute of Chemical Sciences and Engineering, School of Basic Sciences, Ecole Polytechnique Fédérale de Lausanne, CH-1015 Lausanne, Switzerland. 3. Center for Computational Energy Research, Applied Physics, Eindhoven University of Technology, Eindhoven, the Netherlands. 4. Cavendish Laboratory, University of Cambridge, Cambridge, UK. 5. Max-Planck Institute for polymer Research, Ackermannweg 10, Mainz, Germany. 6. Laboratory for Photonics and Interfaces (LPI), Institute of Chemical Sciences and Engineering, School of Basic Sciences, Ecole Polytechnique Fédérale de Lausanne, CH-1015 Lausanne, Switzerland. 7. Laboratory of Magnetic Resonance (LRM), Institute of Chemical Sciences and Engineering, School of Basic Sciences, Ecole Polytechnique Fédérale de Lausanne, CH-1015 Lausanne, Switzerland. 8. Institute of Physical Chemistry, Polish Academy of Sciences, Kasprzaka 44/52, 01-224 Warsaw, Poland. 9. Institute of Microelectronics of Barcelona IMB-CNM, C/del Til·lers., Cjampus Universitat Autònoma de Barcelona (UAB) 08193, Cerdanyola del Vallès (Bellaterra), Barcelona, Spain. 10. ICREA, Pg. Lluís Companys 23, 08010 Barcelona, Catalonia, Spain. Supplementary Figure 1. Performance of perovskite solar cells doped with different organic additives varying chain length. Additive: perovskite = 1:250 (molar ratio). To investigate the impact of the functional groups of organic additives on the efficiency and stability of PSCs, we selected molecules with two main functional groups: phosphonic (– PO(OH)2) and carboxylic (–COOH) acid groups. These two functional groups were selected due to their optimal interaction with the halide perovskite. The molecules under study varied in chain length and shape: Phosphonoacetic acid (H2PP), 3-Phosphonopropionic acid (H3pp), 6- Phosphonopropionic acid (H6pp) and 16-Phosphonopropionic acid (H16pp). We also tested another organic molecule, 4-Methoxybenzylphosphonic acid (4-Methox), with -PO(OH)2 and methoxide (–OCH3) groups. The effect of these additives on efficiency is summarized in Supplementary Figure 1. We show that among these molecules, H3pp leads to maximum efficiencies, slightly lower than those of the reference devices (18.3 % for the reference versus 18 % with H3pp). Based on the results, we have chosen H3pp as a priority for further investigation and optimization. Supplementary Figure 2. Optoelectronic properties as a function of H3pp doping molar ratio. a, power conversion efficiency (PCE), b, short circuit current (Jsc), c, open circuit voltage (Voc) and d, fill factor (FF). H3pp: RbCsMAFA perovskite = 1:500, 1:250, 1:150, 1:75, 1:50, 1:25 and 1:12.5 (molar ratio). Supplementary Figure 3. Optoelectronic parameters of the best device with H3pp: RbCsMAFA perovskite = 1:500 (molar ratio). Supplementary Figure 4. IPCE of the champion H3pp-doped RbCsMAFA perovskite solar cell (1:500) shown in Fig. 1b and Supplementary Figure 3. The integrated Jsc for the device is 23.7 mA/cm2, in agreement with the measured Jsc of 24.5 mA/cm2. Supplementary Figure 5. Statistic optoelectronic properties of the reference and H3pp-doped PSCs (1:500): a, Jsc. b, Voc. c, FF and d, hysteresis index (HI). HI = (PCEReverse-PCEFoward)/PCEReverse. Supplementary Figure 6. Characterization made to complete PSC devices of the type FTO/c- TiO2/m-TiO2/HP/spiro-OMeTAD/Au with (orange) and without (blue) the H3pp additive. (a) Light intensity dependent Voc ; (b) Transient photocurrent, and (c) Open circuit voltage decay. In all cases similar Jsc and Voc with (orange) and without H3pp additive (blue) are observed. Supplementary Figure 7. TRPL of the reference and H3pp doped HP under different light intensity (fluences): a. 1 sun, b, 5 sun, c, 10 sun and d, 100 sun. Supplementary Figure 8. Photoconductivity as a function of pump-probe delay time for reference sample (a) and sample with H3pp additive (1:500) (b), using four different pump fluences (open circles). For both samples, the fluence-dependent dynamics can be described very well by a second order recombination process (only one variable parameter) on top of a flat background (solids lines). Supplementary Figure 9. Operational stability of the reference and H3pp-doped (1:500) CsMAFA PSCs under 1,000 h continuous illumination at 1 sun using white LEDs and MPP tracking. 3 devices for each type were employed. Supplementary Figure 10. XRD patterns of RbCsMAFA perovskite thin films doped with different concentration of H3pp: a, entire patterns with 2θ = 5 - 60°. b, enlarged main perovskite peak of (-111), 2θ = 14°. Supplementary Figure 11. Solid-state nuclear magnetic resonance (NMR). a, Nitrogen-14 solid- state magic angle spinning (MAS) NMR spectra (21.1 T, 298 K and 5 kHz) of (1) bulk mechanochemical α-FAPbI3 and (2) H3pp-doped bulk mechanochemical α-FAPbI3 (1:25 mol/mol). b, Phosphorus-31 MAS NMR spectra at 21.1 T of (3) neat H3pp (298 K, 12 kHz MAS, echo-detected) and H3pp-doped bulk mechanochemical α-FAPbI3 (1 :25 mol/mol) : (4) at 298 K and 20 kHz MAS (echo-detected) and (5) at 100 K and 12 kHz MAS (1H-31P CP MAS), and (6) H3pp-doped bulk mechanochemical RbCsMAFA perovskite (1:5 mol/mol) at 298 K and 20 kHz MAS (echo-detected). To verify the interaction of the H3pp with the perovskite crystal lattice we carried out solid- state NMR measurements. To facilitate the interpretation of the NMR results and ensure that they are unambiguous, we first choose the α-FAPbI3 as a simple perovskite model (FA is the dominant cation in the perovskite applied in this work). It should be noted that analogous solid-state NMR analysis can be carried out for the RbCsFAMA perovskite (Supplementary Fig. 11 b(6)). However, any small variations in the cation ratio or halide ratio, leading to a very slight change in the cubooctahedral symmetry, would result in a change in the 14N envelope making the interpretation of this result ambiguous. Supplementary Figure 11a shows the 14N magic angle spinning (MAS) NMR spectra of bulk α-FAPbI3 (Supplementary Fig. 11 a(1)) and bulk α-FAPbI3 doped with H3pp (Supplementary Fig. 11 a(2)), at 1:25 mol/mol ratio (high doping ratio is required for high quality NMR signal). The 14N spinning sideband (SSB) manifold is narrower in the material doped with H3pp compared to the undoped α-FAPbI3. The width of the 14N SSB manifold is directly related to the symmetry of the cuboctahedral cavity in which FA is reorienting on the picosecond timescale. Narrower SSB manifold corresponds to higher (closer to cubic) symmetry1,2. Thus, doping with H3pp results in the modification of the crystallographic symmetry of α-FAPbI3 in such a way that the resulting symmetry is higher. This result indicates that H3pp interacts with the perovskite lattice at the atomic level and has a structure directing effect similar to the previously described for other mono- and bi-functional molecular modifiers3,4. We further demonstrate this interaction using 31P MAS NMR (Supplementary Fig. 11 b). Pristine H3pp has a 31P chemical shift of 35.4 ppm at 298 K (Supplementary Fig. 11b(3)). In H3pp-doped α-FAPbI3 (1:25 mol/mol) at 298 K (Supplementary Fig. 11 b(4)), there is no unreacted H3pp and a new 31P environment is present at 30.7 ppm, corresponding to H3pp interacting with the perovskite lattice. This chemical environment is well defined as indicated by the narrowness of the peak (FWHM: 223 ± 1 Hz). Upon cooling to 100 K, two types of H3pp species are distinguishable in the H3pp-doped perovskite (Supplementary Fig. 11 b(5)): a narrow peak at 35.3 ppm (FWHM of about 250 Hz) and a broad peak at 32.6 ppm (FWHM of about 3,000 Hz). The chemical shift of the narrow feature at 100 K matches that of neat H3pp (note that neat H3pp is not present in this sample, as evidenced by the spectrum at 298 K). In addition, the longitudinal relaxation rate, T1, of the dipolar coupled protons giving rise to the cross-polarized 31P signal is about 3 s, which is much shorter than the 1 H T1 of the neat H3pp with a value greater than 300 s. The broad peak likely corresponds to H3pp environments that are sensitive to the displacive phase transition of the perovskite lattice (occurring through the octahedral tilt), hence highly disordered at low temperature when the symmetry of the perovskite phase is low. Similar to the model α-FAPbI3 case, there is a new phosphorus environment at 31.6 ppm which corresponds to H3pp interacting with the quadruple cation perovskite lattice on the atomic level (Supplementary Fig. 11 b(6)). Note that this peak is slightly shifted with respect to that found in the model α-FAPbI3 system (30.7 ppm), due to the presence of the bromide and additional A-site cations in the perovskite lattice, which modifies the local environment sensed by the 31P site of H3pp. Note that the signal of unreacted H3pp is also present in this case, since the doping ratio was higher (1:5). Accordingly, we suggest that the narrow component corresponds to H3pp environments which are farther away from the perovskite surface, thus less sensitive to symmetry lowering. Given this data, it is possible to establish that (i) H3pp modifies the perovskite lattice, likely by binding to its surface (ii) H3pp fully reacts with α-FAPbI3 at 1:25 mol/mol ratio (4 mol %). The possible interaction mode is the formation of a surface layer of H3pp in the GBs and on the surface or the formation of perovskite with H3pp in the hollow cation position5. The latter scenario would lead to a significant decrease in the cuboctahedral symmetry (detectable through 14N MAS NMR spectra) as well as leading to new peaks or peak shifts in XRD patterns6 (Supplementary Fig. 10), none of which is observed experimentally. Therefore, we can exclude the hypothesis that H3pp is incorporated into crystal lattice, and conclude that H3pp most likely resides at GBs and on the surface. Molecular H3pp: FAPbI3 interaction modes Supplementary Figure 12. H3pp molecule: FAPbI3 interaction modes on FAI and PbI2 terminated FAPbI3 surfaces, ordered by binding energy from the highest to the lowest. a-b Functional group −PO(OH)2 interacts with FAI terminated surfaces, respectively. c-d Functional group −COOH interacts with FAI terminated surfaces, respectively. f and h Functional −PO(OH)2 and −COOH interact with PbI2 terminated surfaces, respectively. e and g Both −COOH and−PO(OH)2 interact with PbI2 terminated surface. Supplementary Figure 13. Emission spectrum of the white LEDs used in the stability test. Supplementary Table 1. Measured PLQE and estimated QFLS of HP thin films and devices with and without H3pp under one sun. HP H3pp:HP 1:500 Configuration QFLS QFLS PLQE PLQE (eV) (eV) a. Glass/Perovskite HP 1.90 % 1.232 1.94 % 1.232 b. Glass/c-TiO2/m-TiO2/Perovskite HP/Spiro-OMeTAD 0.007 % 1.087 0.004 % 1.074 Quasi-Fermi level splitting (QFLS) determines the theoretical limit of Voc can be obtained for HP and PSCs (Voc,limit = QFLS/q, where q is the elemental charge). The equation to calculate the QFLS from PLQE is7: QFLS =kT ln(PLQE*JG/J0,rad) Where k is the boltzmann constant, T is the temperature, J0,rad is the radiative thermal equilibrium recombination current density in the dark, JG is the generated current density under illumination. For estimation, we assume JG = Jsc = 220 A*m-2, J0,rad = 6.5*10-21 A*m-2 (typical value from the literature), the QFLS values are estimated to be 1.232 eV for the ref and H3pp doped HP, and 1.087eV and 1.074 eV for the ref and H3pp doped PSCs, respectively. Indeed, the same batch of devices of Ref and 1:500 have an average Voc of 1.07 V and 1.06 V, respectively. In sum, the PLQE results confirm the similar non-radiative recombination in the HP thin films and devices with and without H3pp. Supplementary Table 2. Binding energy, vibrational frequency of relevant bonds in H3pp molecule: FAPbI3 systems in Supplementary Figure 12. Configurations Binding energy (eV) Frequency (cm -1) −C=O −P=O FAI termination a 3.13 1756.58 1171.42 b 2.99 1752.63 1207.14 c 2.93 1693.86 1244.14 d 2.90 1733.14 1244.52 PbI2 termination e 2.54 1764.79 1182.36 f 1.91 1753.00 1097.63 g 1.84 1692.72 1276.74 h 1.59 1642.62 1234.21 Note I: Spin-polarized DFT framework for charged defect formation energy calculations. We calculated the formation energies of FA+ and I- at the FAI terminated surface of the H3pp - modified FAPbI3, respectively. As a reference, we also calculate the same formation energies for the pristine FAI terminated FAPbI3. The defect formation energy is given as, DEF ( X q )  E ( X q )  E (bulk )   ni i  q( EF  EVBM  V ) i (0.1) q where E ( X ) is the total energy of the defect system with charge q , E (bulk ) is the total energy of the pristine system, and n and  are the number and the chemical potentials of the species added to or subtracted from the perfect system in order to form the defect. The last term is the energy associated with the exchange of charges with the electrons’ reservoir (the Fermi level of the system E F ), referenced to the valence band maximum ( EVBM ) of the pristine system, followed by a correction for the electrostatic potential shift V induced by the defect. The last term is assumed to be the same for H3pp-modofied and -unmodified FAPbI3 because similar Quasi Fermi level splitting (QFLS) for the two perovskites is found in photoluminescence quantum efficiency (PLQE) observations (Supplementary Table 1). We define the relative formation energy as in Ref. 8. The suppression of FA+ and I- vacancy formation due to the introduction of H3pp molecule is quantitatively evaluated by, Evac  DEF ( X q , H 3 pp )  DEF ( X q , ref ) (0.2) q where DEF ( X , H 3 pp) is the formation energy of a FA or I vacancy in the presence of H3pp q molecule and DEF ( X , ref ) the one of the unmodified FAI-terminated surface. References 1. Kubicki, D. J. et al. Cation dynamics in mixed-cation (MA)x(FA)1–xPbI3 hybrid perovskites from solid- state NMR. J. Am. Chem. Soc. 139, 10055-10061 (2017). 2. Kubicki, D. J. et al. Formation of stable mixed guanidinium–methylammonium phases with exceptionally long carrier lifetimes for high-efficiency lead iodide-based perovskite photovoltaics. J. Am. Chem. Soc. 140, 3345-3351 (2018). 3. Bi, D. et al. Multifunctional molecular modulators for perovskite solar cells with over 20% efficiency and high operational stability. Nat. Commun. 9, 4482 (2018). 4. Tavakoli, M. M. et al. Addition of adamantylammonium iodide to hole transport layers enables highly efficient and electroluminescent perovskite solar cells. Energy Environ. Sci. 11, 3310-3320 (2018). 5. Spanopoulos, I. et al. Unraveling the chemical nature of the 3D “hollow” hybrid halide perovskites. J. Am. Chem. Soc. 140, 5728-5742 (2018). 6. Mei, A. et al. A hole-conductor–free, fully printable mesoscopic perovskite solar cell with high stability. Science 345, 295-298 (2014). 7. Stolterfoht, M. et al. The impact of energy alignment and interfacial recombination on the internal and external open-circuit voltage of perovskite solar cells. Energy Environ. Sci. 12, 2778-2788, doi:10.1039/C9EE02020A (2019). 8. Li, N., Tao, S., Chen, Y. et al. Cation and anion immobilization through chemical bonding enhancement with fluorides for stable halide perovskite solar cells. Nat. Energy 4, 408–415 (2019). https://doi.org/10.1038/s41560-019-0382-6. Photovoltaic Cell Checklist Standardized data reporting for photovoltaic cells Revision 1.0, Last updated: Sept 11, 2020 If a key point of your paper is the performance of a photovoltaic cell, complete the below form »» Authors are encouraged to include this completed document as supplemental information at the time of submission. Not doing so may cause delays if your paper is selected for peer review. »» Please confirm that each item in the list is present in the manuscript or supplemental information by checking the corresponding box. For any items that are not present, please provide details in the text box for the editorial team and reviewers to consider. Performance reporting Explain exception if line item is not present in manuscript Current-voltage characteristics plotted in ✔ both forward and reverse directions confirm Fig. 1b Current-voltage scan speed and dwell ✔ time confirm 40 mV/s, 1 min Details of device preconditioning, if any, ✔ prior to testing confirm No preconditioning *for example light soaking or holding cell at a bias Testing environmental conditions ✔ *such as atmosphere, temperature, and humidity confirm Room temperature, in air, RH 50% Statistical breakdown of performance ✔ across multiple devices confirm more than 10 devices, also see Fig. 1c *including number of individual substrates tested Performance of any control devices are ✔ benchmarked to literature confirm Details of antireflection coating, if used ✔ confirm No Cell area Device contact area ✔ confirm 0.025 cm2 Device illuminated area during testing ✔ confirm 0.016 cm2 How contact and illumination area were ✔ determined, including mask/aperture confirm a mask with a square area of 0.016 cm2 details Performance reporting at 1 cm2 cell ✔ *highly desirable but not mandatory confirm No Device stability Evolution of efficiency or current density ✔ (Jsc) at maximum power point confirm inset of Supplementary Fig. 3 *at least 100 seconds is recommended Atmosphere, temperature, humidity, bias, ✔ illumination, and storage conditions for confirm N2, 25 C, MPP, 1 sun long-term testing Encapsulation details, if used ✔ confirm No Supporting characterization External quantum efficiency (EQE) of key ✔ representative devices confirm Supplementary Fig. 4 Quantification of mismatch between ✔ integrated Jsc from EQE with Jsc from confirm Supplementary Fig. 4 operational illumination source Total reflectance or absorption spectra of ✔ cell illuminated from front side confirm No Dark current-voltage characteristics ✔ shown on logarithmic current axis in both confirm No sweep directions For tandem devices: EQE, bias voltage, ✔ and bias illumination of each subcell confirm No Performance verification *highly desirable for performance claims near or exceeding state-of-art Certification or other independent ✔ verification of performance results confirm No Calibration of illumination source and reference cell Model of light source, reference cell, and ✔ other power sensors used in testing halogen lamp, Si solar cell calibrated by Fraunhofer for IV test, and white LEDs for confirm stability test. Date when reference cell was last ✔ calbriated and certified confirm three months before IV measurements Spectral mismatch between reference cell ✔ and testing cells confirm 0.99 Spectra of solar simulator used for ✔ measurement, including any additional confirm No UV filter filters used, and date it was measured For further reading, see: 1. Solar Cells Reporting Summary. Nature Research. https://www.nature.com/documents/nr-photovoltaic-reporting.pdf 2. Reese et. al. Reliably Measuring the Performance of Emerging Photovoltaic Solar Cells. Nanostructured Materials for Type III Photovoltaics, 1–32 (2017). 3. Wang et. al. Reliable Measurement of Perovskite Solar Cells. Adv. Mater. 31, 1803231 (2019). 4. Timmreck et. al. Characterization of tandem organic solar cells. Nature Photon. 9, 478–479 (2015). 5. Kiermasch et. al. Effects of Masking on Open-Circuit Voltage and Fill Factor in Solar Cells. Joule 3, 16–26 (2019). 6. Fell. Standardising current–voltage measurements for metastable solar cells. JPhys Energy. 2, (2019). 7. ASTM. Committee E44 on Solar, Geothermal and Other Alternative Energy Sources. 8. IEC. TC 82 Solar photovoltaic energy systems. I verify that this form is completed accurately in agreement with all co-authors, to the best of my knowledge ✔ Submitting author name: Monica Lira-Cantu For general queries or feedback regarding this form please email joule@cell.com