Nature Materials: Anion optimization of bifunctional surface passivation for perovskite solar cells

• 1. Background introduction
  Pseudohalide (PH) anion engineering has received widespread attention as a surface passivation strategy in perovskite-based optoelectronics; however, so far, defect passivation by pH anions has been insufficient, creating deep impurity states. So far, it has been difficult to explore all candidate molecular families because the chemical space of pH anions is too large (>106 molecules).
  

  
  

  
  

  2. Research problems

  This study created a machine learning workflow that leverages full density functional theory calculations to train models to speed up the development process. This physics-based machine learning model allows us to pinpoint promising molecules with head groups that prevent lattice distortion and the formation of anti-site defects and tail groups optimized for strong attachment on the surface. Through experiments, this study found that sodium thioglycolate is the most effective passivating agent. Using this strategy, the inverted perovskite solar cell achieved a power conversion efficiency of 24.56% and an open-circuit voltage as high as 1.19 volts (National Renewable Energy Laboratory-certified quasi-steady-state efficiency of 24.04%). The packaged device maintained 96% of the initial power conversion energy during 900 hours of single-solar operation at the maximum power point.
  

  
  
  

• 

• 
  

  Figure 1 | Workflow highlights for identifying candidate pH anions as passivators to improve photovoltaic performance in concentrating solar cells: 1. To do this, this study used density functional theory (DFT) on a subset of chemical space, A machine learning (ML) model was thus trained (Figure 1a). First, experimentally reasonable candidate compounds are screened in pH space. A multi-step screening funnel considers charge state (Level 1), molecular weight (Level 2), availability of three-dimensional structure (Level 3), molecular radius (Level 4), presence of Na or K salts (Level 5 ) and the availability of salt (Tier 6). In this study, only 168 pH anions were selected from 5 million molecules extracted from the PubChem database. 2. This study used FA0.75MA0.25PbI3 as a model system to study the interaction between PH anions and the perovskite surface (Figure 1b). PH anions can interact with the perovskite surface either by substitution (a PH anion replaces an X(I-) atom) or by adsorption (binding with an undercoordinated B2+ (Pb2+) cation). According to the formation energy (ΔEformation). (Fig. 1c), this study found that substitution is thermodynamically more favorable than adsorption in most I chemical potential (ΔμI) regions. This is attributed to the formation of more ionic bonds between Pb2+ and the electron-rich groups of the PH anion, and the formation of more hydrogen bonds between the NH3+ groups of FA+/MA+ and the PH anion. The charge density difference reveals the charge transfer of PH anions between the surface and the perovskite plate, indicating that PH anions can attract electrons from neighboring Pb atoms and MA/FA atoms.

  
  

• 

• 
  

  Figure 2 | Determining the ranking and chemical effects of physical features that dominate the binding energy performance of candidate pH anions. Key points: 1. Based on the understanding of the passivation mechanism of pH anions, this study performed DFT calculations to estimate the binding energy (Eb) of pH anions to the perovskite surface. As shown in Figure 2a, the Eb of several new pH anions is even higher than the 16 anions studied in earlier literature. The calculated Eb trend in FA0.75MA0.25PbI3 seems to extend well to other perovskite compositions such as FA0.75Cs0.25PbI3 (normal band gap), FA0.75MA0.25Pb0.5Sn0.5I3 (narrow band gap) and FA0.75Cs0.25Pb(I0.625Br0.375)3 (wide bandgap). Since higher Eb values ​​indicate stronger binding strength to iodine vacancies VI, this study predicts that these candidate substances will enhance the passivation effect. 2. This study then attempts to develop a physically informed machine learning (ML) model to study how the molecular structure of pH anions modulates the strength of their interaction with the perovskite surface. The screened PH anions have a variety of electron-rich functional groups, including R-SO3-, R-SO2-, R-CO2-, R-COS-, R-S2O2-, R-CS2-, R-BF3-, PO3- , R-PO2-, R-PHO3-, R-S-, R-O-. Jeong et al. report that the pH anion formate (HCOO-) has a binding energy of 3.1 eV, resulting in n-I-p (negative-endo-positive) PSCs with record performance, while the chloride (Cl-) binding energy is 2.98 eV. In view of this, this study classified 267 pH anions (including 168 p-h anions screened in this study (Fig. 1a)) into high/low Eb based on 3 eV: therefore pH anions were either labeled as high Eb ( 201 anions), or labeled as low Eb (66 anions) and output as ML. As shown in Figure 2b, this study initially used 19 features of pH anions as ML input. 3. After two rounds of ML training, four main features affecting Eb classification were determined, including the number of oxygen atoms (num_O), topological polar surface area (TPSA), number of hydrogen bond acceptors (HBA) and the highest occupied molecular orbital level ( HOMO). Using these four features, this study achieved an area under the receiver operating characteristic (ROC) curve (AUC) score of 0.87 and an accuracy score of 0.84 in the random forest model (Figure 2c). In the regression model, features with positive coefficients favor positive labels (Fig. 2c). Therefore, this study provides a set of guiding principles for the determination of binding energy: the more num_O of the PH anion, the larger the TPSA, the more HBA, and the lower the HOMO level, the stronger the binding force to the perovskite surface.

  
  
  

• 

• 
  

  Figure 3 | Computational study points for bifunctional candidate ligands: 1. This study also attempted to prevent deep impurity states caused by the pH treatment itself, as they would increase non-radiative losses. This prompted us to search for safe functional groups: in the higher Eb levels (Eb > 3 eV), 24 potential pH anions were found (Figs. 2a and 3d), which do not show any significant differences in the band, regardless of the calculation method used. Local states are generated near the edge. 2. The emerging candidate molecules prompted us to study the impact of functional groups on the formation of IPb anti-site defects (Figure 3a). It is known that IPb anti-site defects can lead to deep sites within the (MA/FA)PbI3 perovskite band gap. . This study found that the SO3- functional group will cause greater surface structural distortion, thereby inhibiting the passivation of negatively charged IPb anti-site defects, while CO2- will prevent the formation of IPb, but as mentioned above, its surface binding strength is relatively low. Low. 3. This study also attempts to calculate the anion migration barrier (ΔEa) on the perovskite surface, Because halide migration is one of the causes of current-voltage hysteresis and decomposition of perovskite films. As shown in Figure 3c, the ΔEa of I- (vacancy-mediated) is 0.23 eV (average), which is consistent with previous calculations. This study found that the ΔEa of Br-, Cl-, and F- increased by 0.27 eV, 0.32 eV, and 0.37 eV, respectively, which is related to the enhanced electronegativity and binding strength to the perovskite surface (Figure 2a). The optimal bifunctional pH anion (ligand 14) further increases ΔEa to 0.43 eV, which is a result of enhanced binding strength and rotational barriers of the additional pH anion, FA+/MA+ molecules, which are not present in the halide (I -, Br-, Cl-) during the migration process.

  
  

• 

• 
  

  Figure 4 | Comparison points between experimental research on the performance of pH anion treatment equipment and related control groups: 1. In the experiment, this study selected five bifunctional pH anions with higher Eb values ​​for further research. In this study, the pH anion salt was dissolved in isopropyl alcohol (IPA) solution, spin-coated on the perovskite film at a concentration of 0.5 mg/ml, and then annealed at 100 °C. Then, inverted PSC devices without pH anion treatment and with pH anion treatment were fabricated in this study (Fig. 4a). This study uses Cs0.05FA0.9MA0.05Pb(I0.95Br0.05)3 perovskite with a band gap of 1.55 eV as the absorber. These PSCs have an inverted device structure of ITO/NiOx/Me-4PACz/perovskite/C60/BCP/Ag, where ITO is indium tin oxide and Me-4PACz is [4-(3,6-dimethyl-9H-carb Azol-9-yl)butyl]phosphonic acid, C60 is fullerene, and BCP is bathophore. In PSCs with an inverted (p-I-n) device structure, due to the large amount of interfacial non-radiative recombination, Therefore, the interface between the perovskite layer and the electron transport layer (ETL) has been proven to be the most critical factor affecting device performance. This prompted this study to focus on the use of a pH anionic passivation layer at the perovskite/ETL interface. 2. Figure 4b shows the current density-voltage (J-V) curves of the control device and the device treated with five pH anions under forward and reverse scanning conditions. The graph shows that sodium thioglycolate (ST) performed best, but all five pH anions showed improved performance compared to the control. This is due to the optimal bifunctional passivation effect of thioglycolate. Both control and pH anion-treated devices exhibited lower hysteresis. 3. Figure 4c shows the use of five different pH anion treatments (ST (ligand 14); sodium chlorate (SC) (ligand 2); potassium bicarbonate (PB) (ligand 7); sodium glycolate (SH) ) (ligand 13); device performance statistics of monosodium methylphosphonate (MMP) (ligand 12)). This study found that the volatile organic compounds (VOC) and PCE of these five pH anion treatment devices decreased in order: ST was the best, SC was the worst, and SH, MMP and PB were in between.

  
  

• 

• 
  

  Figure 5 | Experimental characterization and device stability of ST-treated perovskite films Key points: 1. Next, this study characterized the morphology and crystallinity of the perovskite films to reveal the relationship between perovskite and PH anion treatment interaction between. Top-down and cross-sectional scanning electron microscopy (SEM) images show that the film morphology remains unchanged after pH anion treatment. The pH anion treatment did not affect the overall crystallinity and perovskite phase, with no false peaks and no associated peak shifts, broadening or intensity differences, as can be seen from the index films and simulated X-ray diffraction (XRD) patterns. prove. In addition, ST and MMP treatment can also reduce excess PbI2, which is also beneficial to stability (Fig. 5a). 2. In order to determine whether the best-performing PH anion (thioacetate) exists after post-treatment, this study used X-ray photoelectron spectroscopy (XPS), whose O1s peak (Figure 5b) and S 2p peak (Figure 5c) It shows that the pH anion is anchored on the surface of the perovskite film. After ST treatment, the binding energies of Pb 4f5/2 and Pb 4f7/2 orbitals shifted by 0.24 eV and 0.34 eV, respectively, indicating a strong interaction between the pH anions and the perovskite surface (Fig. 5d).

  
  

  Summary and outlook

  This study demonstrates that continued progress in the integration of computation, ML, and experiment provides further potential for the discovery of molecular strategies to improve the performance of optoelectronics, including photovoltaics, and related light-emitting devices.

  
  

  Paper link:

  Xu, J., Chen, H., Grater, L. et al. Anion optimization for bifunctional surface passivation in perovskite solar cells. Nat. Mater. (2023).

  
  

  https://doi.org/10.1038/s41563-023-01705-y

  https://www.nature.com/articles/s41563-023-01705-y