In my work on inverted perovskite solar cells, I have become increasingly convinced that the gap between laboratory-scale record efficiencies and the performance predicted by the Shockley–Queisser limit is dominated not by the intrinsic absorption strength of the halide perovskite absorber, but by defect-mediated nonradiative recombination. The mixed-cation, mixed-halide perovskite Cs0.05MA0.10FA0.85PbI3 that I use as the light absorber possesses an absorption onset that is almost ideally matched to the solar spectrum, carrier diffusion lengths exceeding one micrometer, and ambipolar transport. Yet the polycrystalline nature of the film inevitably introduces grain boundaries, residual lead iodide, undercoordinated Pb2+ centers, and halogen vacancies that act simultaneously as deep-level traps, as shunting pathways, and as channels for moisture and thermal degradation. I therefore set out to design a single molecular additive that could address all of these failure modes concurrently rather than sequentially.
My central hypothesis was that a hydrochloride salt carrying both a pyridine nitrogen and a hydrazine moiety would engage the perovskite lattice through several distinct chemical handles at once: Lewis acid–base coordination to undercoordinated Pb2+, hydrogen bonding to iodide, electrostatic interaction through the protonated ammonium-type site, and halide compensation through the chloride counterion. To test this hypothesis I selected 4-hydrazinopyridine hydrochloride (4HPCl) and introduced it directly into the perovskite precursor ink, so that the additive would be present during nucleation and crystal growth rather than only at the post-deposition surface. The additive concentration was varied across 0, 0.25, 0.50, and 1.00 mg·mL−1, and the resulting films and devices were characterized with a full suite of structural, optical, chemical, and electrical probes.
In this article I describe the fabrication protocol, the crystallization and morphology optimization, the optoelectronic evidence for suppressed nonradiative recombination, the X-ray photoelectron spectroscopy results that establish the chemical nature of the passivation, the space-charge-limited current and impedance analysis that quantify the reduction in trap density and the enhancement of the built-in potential, and finally the photovoltaic and thermal-stability performance of the resulting perovskite solar cells. The optimized device reaches a power conversion efficiency of 25.31% with an open-circuit voltage of 1.166 V and a fill factor of 83.40%, and, more importantly from my perspective, retains roughly 90% of its initial efficiency after 900 h of thermal aging at 65 °C in a nitrogen atmosphere.
1. Molecular Design Rationale
The 4HPCl molecule is an aromatic heterocyclic organic amine salt. Its structure combines a positively charged pyridinium-type site, a hydrazine group (–NH–NH2) that is strongly electron-donating, and a chloride counterion. I reasoned that this combination would produce a multifunctional passivator whose interaction with the perovskite is not limited to a single binding motif. The pyridine nitrogen can donate its lone pair to vacant lead orbitals; the hydrazine terminus can participate in both coordination and hydrogen bonding; and the chloride anion can occupy iodine vacancies or reorganize the local halide coordination environment at grain boundaries. Because the molecule is small and highly soluble in the N,N-dimethylformamide/dimethyl sulfoxide solvent system, it disperses homogeneously throughout the wet film during spin coating and remains available at the grain boundaries after crystallization.
I summarize the principal interaction pathways that I expected to be operative in Table 1.
| Functional motif in 4HPCl | Proposed interaction with perovskite | Expected consequence |
|---|---|---|
| Pyridine N lone pair | Lewis acid–base coordination to undercoordinated Pb2+ | Filling of deep-level Pb-related traps |
| Hydrazine –NH–NH2 | Coordination and hydrogen bonding to iodide and to the [PbI6]4− framework | Suppression of iodide vacancy defects |
| Protonated N site | Electrostatic interaction with negatively charged halide-rich surfaces | Stabilization of grain boundaries |
| Chloride counterion | Halide-site compensation and local coordination reconstruction | Reduction of residual PbI2 and halide disorder |
The working hypothesis can be expressed as a multi-site equilibrium in which the additive A competes with native defects for binding sites on the perovskite surface:
$$A + \mathrm{Pb^{2+}_{uc}} \rightleftharpoons A\text{–}\mathrm{Pb^{2+}_{coord}}$$
$$A + \mathrm{V_I^{+}} \rightleftharpoons A\text{–}\mathrm{I_{bound}}$$
Here Pb2+uc denotes an undercoordinated lead site and VI+ denotes an iodine vacancy. The extent to which these equilibria are driven toward the passivated products depends on the additive concentration, and this concentration dependence is precisely what I observe experimentally.
2. Fabrication of the Perovskite Solar Cells
The inverted architecture that I adopted was glass/ITO/NiOx/Ph-4PACz/perovskite/PEAI/C60/BCP/Ag. I cleaned indium tin oxide substrates sequentially in detergent, ultrapure water, acetone, and anhydrous ethanol, each for 15 min under ultrasonication, dried them under nitrogen, and activated the surface with ozone for 15 min. A 10 mg·mL−1 NiOx nanoparticle dispersion in ultrapure water was ultrasonicated and shaken overnight, then spin-coated at 3000 rpm for 30 s and annealed at 120 °C for 20 min. A 0.5 mg·mL−1 ethanol solution of the phosphonic acid hole-transport molecule was deposited at 3000 rpm for 30 s and annealed at 100 °C for 10 min.
For the perovskite layer I dissolved CsI, MAI, FAI, and PbI2 in a 5:1 v/v mixture of DMF and DMSO to obtain the target composition Cs0.05MA0.10FA0.85PbI3, with an additional 15 mol% MACl and 5 mol% excess PbI2, giving a total concentration of 1.6 mol·L−1. The solution was shaken for 4 h and filtered through a 0.22 μm PTFE membrane. For the modified samples, an appropriate volume of the filtered solution was transferred into a vial containing a weighed amount of 4HPCl and shaken for a further 2 h. The films were spin-coated in a two-step program (1000 rpm for 10 s, then 5000 rpm for 35 s) with 220 μL of ethyl acetate dripped as antisolvent during the final 15 s of the second step, followed by annealing at 100 °C for 30 min. A PEAI isopropanol solution (1 mg·mL−1, 60 μL, 5000 rpm, 30 s) was used as a surface passivation treatment. Finally, C60 (25 nm), BCP (7 nm), and silver (100 nm) were deposited by thermal evaporation.
Because every sample in this study was deposited on the same type of substrate with identical spin parameters, antisolvent volume, and antisolvent timing, I treat the film thickness as effectively constant across the control and modified groups. This allows me to attribute the observed electrical differences to changes in the perovskite layer itself rather than to geometric variations.
3. Crystallization and Morphology Optimization
The X-ray diffraction patterns of the control film show a distinct reflection near 12.7° that corresponds to residual PbI2. This reflection is a well-known signature of incomplete conversion or of excess lead iodide that has segregated at grain boundaries. After introducing 4HPCl, this reflection disappears from the patterns, which I interpret as evidence that the additive promotes more complete reaction of the lead precursor and suppresses the accumulation of unreacted PbI2. Quantitatively, the ratio of the PbI2 reflection intensity to that of the perovskite (100) reflection falls from 0.7 in the control to values below the detection limit in the treated films.
The full width at half maximum of the perovskite reflections provides a complementary measure of crystalline quality. I compared the three additive concentrations and found that 0.25 mg·mL−1 produces a full width at half maximum comparable to that of the control, that 1.00 mg·mL−1 produces a value even larger than the control, and that 0.50 mg·mL−1 produces the narrowest reflection. Since a narrower reflection corresponds to a larger coherently diffracting domain and to reduced microstrain, I identified 0.50 mg·mL−1 as the optimum additive concentration and designated this formulation as the target sample for all subsequent analysis. I collect these structural observations in Table 2.
| Additive concentration (mg·mL−1) | Residual PbI2 reflection at 12.7° | Relative FWHM of perovskite reflection | Interpretation |
|---|---|---|---|
| 0 (control) | Clearly present | Reference value | Incomplete conversion, residual PbI2 |
| 0.25 | Absent | Comparable to control | Partial improvement, limited effect |
| 0.50 | Absent | Narrowest | Optimal crystalline quality |
| 1.00 | Absent | Wider than control | Excess additive perturbs growth |
Scanning electron microscopy confirms this picture. The control film consists of relatively small grains with an average size of approximately 255 nm, exhibits noticeable surface roughness, and shows numerous bright features at the grain boundaries that I assign to unreacted PbI2. Such morphological features are detrimental because they create interfacial recombination centers and impede charge extraction. After modification with 0.50 mg·mL−1 4HPCl, the average grain size increases to approximately 701 nm, the surface becomes markedly smoother, and the density of grain-boundary PbI2 residues drops substantially. The simultaneous reduction in grain-boundary area and in residual lead iodide is consistent with the disappearance of the 12.7° reflection in the diffraction data.
The grain-size enlargement can be understood in terms of the classical nucleation and growth framework. If the additive reduces the steady-state nucleation rate J while leaving the growth velocity v approximately unchanged, then the final grain size d scales as
$$d \propto \left(\frac{v}{J}\right)^{1/4}$$
A reduction in J therefore translates directly into larger grains. I attribute the reduced nucleation rate to the formation of soluble intermediate complexes between 4HPCl and the lead iodide species in the precursor, which raise the effective supersaturation barrier and delay the onset of nucleation, allowing the existing nuclei to grow to a larger size before impingement.
4. Optical Properties and Nonradiative Recombination
Ultraviolet–visible absorption spectroscopy shows that the 4HPCl-modified film absorbs slightly more strongly across the visible range than the control film, which I attribute to the improved film quality and reduced scattering from rough surfaces and secondary phases. The more important observation is in the photoluminescence data. The steady-state photoluminescence intensity of the modified film is substantially higher than that of the control, indicating that the density of nonradiative recombination channels has been reduced.
To quantify the carrier dynamics I fitted the time-resolved photoluminescence decays with a bi-exponential model:
$$Y(t) = A + B_1 e^{-t/\tau_1} + B_2 e^{-t/\tau_2}$$
where A is a baseline constant, and the two exponential terms represent a fast decay component associated with trap-mediated recombination at the surface or at grain boundaries and a slow decay component associated with bulk radiative recombination. I computed the average carrier lifetime according to
$$\tau_{ave} = \frac{A_1 \tau_1^2 + A_2 \tau_2^2}{A_1 \tau_1 + A_2 \tau_2}$$
The control film yields an average lifetime of 403.77 ns, whereas the 4HPCl-modified film yields 1056.12 ns, an increase of more than a factor of two and a half. In parallel, the photoluminescence quantum yield of the modified film is more than twice that of the control. These results together point to a substantial reduction in nonradiative recombination losses, which is exactly what I expect from a passivator that removes deep-level traps from grain boundaries and surfaces.
I also performed two-dimensional photoluminescence intensity and lifetime mapping to evaluate spatial uniformity. The control film exhibits a heterogeneous intensity distribution with numerous dark regions that correspond to locally high densities of nonradiative recombination centers, and its average mapped lifetime is 27.16 ns with a broad, skewed distribution. The modified film exhibits a markedly brighter and more uniform intensity map, and its average mapped lifetime rises to 48.07 ns with a narrower distribution. Moreover, in the control film the regions of highest photoluminescence intensity do not coincide with the regions of longest lifetime, whereas in the modified film the two maps are well correlated. This spatial correlation is a strong indicator that the passivation is uniform across the film rather than concentrated in isolated regions. I summarize the optical and carrier-dynamic results in Table 3.
| Parameter | Control | 4HPCl-modified (0.50 mg·mL−1) |
|---|---|---|
| Average TRPL lifetime (ns) | 403.77 | 1056.12 |
| Average mapped PL lifetime (ns) | 27.16 | 48.07 |
| Relative PLQY | 1.0 | > 2.0 |
| PL uniformity | Heterogeneous, dark regions | Uniform, spatially correlated with lifetime |
5. Chemical Evidence for Multi-Site Passivation
X-ray photoelectron spectroscopy provides the most direct evidence for the chemical interaction between 4HPCl and the perovskite. After modification, the Pb 4f core levels shift noticeably, which I attribute to coordination between the nitrogen-containing sites of the additive and undercoordinated Pb2+. The I 3d levels also shift, and the direction of the shift toward lower binding energy indicates an increase in local electron density around the iodide ions, consistent with hydrogen bonding or electrostatic interaction that stabilizes the halide sublattice and suppresses iodine vacancy defects. The N 1s signal shifts from 400.39 eV to 400.25 eV, confirming that the chemical environment of the nitrogen atoms changes upon incorporation into the film.
Taken together, these core-level shifts establish that 4HPCl is not merely physically adsorbed but is chemically bound to the perovskite through multiple sites. The chloride counterion contributes an additional passivation channel by occupying halide vacancies and by participating in the local reconstruction of the halide coordination environment. The multifunctional nature of the molecule is therefore essential: a single-binding-site additive could not simultaneously address lead-related and halide-related defects. I summarize the observed core-level shifts and their assignments in Table 4.
| Core level | Observed change after 4HPCl modification | Chemical interpretation |
|---|---|---|
| Pb 4f | Clear binding-energy shift | Coordination of N sites to undercoordinated Pb2+ |
| I 3d | Shift toward lower binding energy | Increased local electron density, iodide vacancy suppression |
| N 1s | 400.39 eV → 400.25 eV | Changed nitrogen chemical environment, hydrogen bonding and coordination |
6. Charge Transport, Trap Density, and Built-in Potential
Dark current–voltage measurements reveal that the leakage current density of the control device is 7.14 × 10−6 mA·cm−2, while that of the 4HPCl-modified device is 1.16 × 10−6 mA·cm−2, a reduction by roughly a factor of six. A lower leakage current indicates fewer shunt pathways and reduced nonradiative recombination at the interfaces and in the bulk, both of which are prerequisites for a high open-circuit voltage and a high fill factor.
I then fabricated electron-only devices with the structure ITO/SnOx/perovskite/C60/BCP/Ag and measured the space-charge-limited current to extract the trap density. The trap-filled limit voltage decreases from 0.51 V in the control to 0.33 V in the modified device. The trap density is obtained from
$$N_{trap} = \frac{2 \varepsilon_0 \varepsilon_r V_{TFL}}{e L^2}$$
where ε0 is the vacuum permittivity, εr is the relative permittivity of the perovskite, e is the elementary charge, and L is the film thickness. Using the same geometric and dielectric parameters for both devices, the trap density falls from 4.68 × 1015 cm−3 to 3.02 × 1015 cm−3. This reduction of approximately 35% is entirely consistent with the longer carrier lifetimes and the higher photoluminescence quantum yield that I measured optically.
Electrochemical impedance spectroscopy provides complementary information about the transport and recombination resistances. The series resistance decreases from 14.19 Ω in the control to 11.61 Ω in the modified device, while the recombination resistance increases from 8086 Ω to 16621 Ω. The decrease in series resistance reflects improved charge extraction and higher effective carrier mobility, whereas the increase in recombination resistance reflects the suppression of trap-mediated recombination. The two effects are synergistic: better extraction reduces the steady-state carrier density that would otherwise be available for recombination, and fewer traps reduce the recombination rate at any given carrier density.
Mott–Schottky analysis shows that the built-in potential increases from 0.66 V in the control to 0.87 V in the modified device. A larger built-in potential provides a stronger driving force for the separation and extraction of photogenerated carriers. Because the open-circuit voltage of a perovskite solar cell is bounded above by the built-in potential minus the recombination-induced loss, the increase in built-in potential directly supports the higher open-circuit voltage that I observe. I collect all of the electrical parameters in Table 5.
| Electrical parameter | Control | 4HPCl-modified | Physical meaning |
|---|---|---|---|
| Dark leakage current density (mA·cm−2) | 7.14 × 10−6 | 1.16 × 10−6 | Shunt pathways and interfacial recombination |
| Trap-filled limit voltage (V) | 0.51 | 0.33 | Onset of trap filling |
| Trap density (cm−3) | 4.68 × 1015 | 3.02 × 1015 | Deep-level defect concentration |
| Series resistance (Ω) | 14.19 | 11.61 | Ohmic and contact losses |
| Recombination resistance (Ω) | 8086 | 16621 | Resistance to charge recombination |
| Built-in potential (V) | 0.66 | 0.87 | Driving force for carrier separation |
7. Photovoltaic Performance
The photovoltaic parameters of the devices were measured under standard AM 1.5G illumination at 100 mW·cm−2 with a mask defining an active area of 0.049 cm2. I first screened the four additive concentrations to identify the optimum. All modified devices outperformed the control, but the concentration dependence was non-monotonic, with 0.50 mg·mL−1 giving the best overall result. I summarize the screening outcome in Table 6.
| 4HPCl concentration (mg·mL−1) | VOC (V) | JSC (mA·cm−2) | FF (%) | PCE (%) |
|---|---|---|---|---|
| 0 (control) | 1.105 | 25.94 | 78.06 | 22.38 |
| 0.25 | Improved over control | Slightly improved | Improved | Moderate gain |
| 0.50 | 1.166 | 26.04 | 83.40 | 25.31 |
| 1.00 | Lower than optimum | Lower than optimum | Lower than optimum | Below optimum |
For the champion device the efficiency, open-circuit voltage, short-circuit current density, and fill factor are 25.31%, 1.166 V, 26.04 mA·cm−2, and 83.40%, respectively. The corresponding control values are 22.38%, 1.105 V, 25.94 mA·cm−2, and 78.06%. The efficiency gain is therefore 2.93 percentage points in absolute terms, corresponding to a relative improvement of approximately 13%. I note that the largest relative gains are in the open-circuit voltage and the fill factor rather than in the short-circuit current density, which is exactly the pattern expected when the dominant loss mechanism being addressed is nonradiative recombination rather than optical absorption.
The relationship between the three photovoltaic parameters and the efficiency is, of course,
$$\mathrm{PCE} = \frac{V_{OC} \, J_{SC} \, \mathrm{FF}}{P_{in}}$$
with Pin = 100 mW·cm−2. Substituting the champion values gives
$$\mathrm{PCE} = \frac{1.166 \times 26.04 \times 0.8340}{100} \approx 25.31\% $$
which is internally consistent with the measured value.
Hysteresis is an important reliability metric for perovskite solar cells because it reflects the presence of mobile ions and interfacial charge accumulation. I define the hysteresis index as
$$\mathrm{HI} = \frac{\mathrm{PCE}_{reverse} – \mathrm{PCE}_{forward}}{\mathrm{PCE}_{reverse}}$$
The control device gives a reverse-scan efficiency of 22.38% and a forward-scan efficiency of 20.87%, corresponding to a hysteresis index of 6.75%. The modified device gives 25.31% and 24.82%, corresponding to a hysteresis index of only 1.93%. The suppression of hysteresis indicates that ionic migration and interfacial charge trapping have been substantially reduced by the passivation treatment, which is consistent with the reduced trap density extracted from the space-charge-limited current measurements.
Steady-state output measurements at the maximum power point confirm the current–voltage results. The modified device held at 1.02 V delivers a steady-state current density of 24.74 mA·cm−2, corresponding to a steady-state efficiency of 25.23%. The control device held at 0.94 V delivers 23.72 mA·cm−2 and 22.25%. The close agreement between the steady-state and the scanned efficiencies rules out significant transient artifacts in the current–voltage curves.
The external quantum efficiency spectra show that the modified device has a slightly higher integrated current density, rising from 25.12 mA·cm−2 in the control to 25.27 mA·cm−2 in the modified device. This small increase is consistent with the small increase in short-circuit current density and confirms that the primary benefit of the additive is in the voltage and fill-factor channels rather than in photocurrent generation. I collect the full device performance comparison in Table 7.
| Device metric | Control | 4HPCl-modified |
|---|---|---|
| Reverse-scan PCE (%) | 22.38 | 25.31 |
| Forward-scan PCE (%) | 20.87 | 24.82 |
| Hysteresis index (%) | 6.75 | 1.93 |
| Steady-state current density (mA·cm−2) | 23.72 | 24.74 |
| Steady-state PCE (%) | 22.25 | 25.23 |
| Integrated JSC from EQE (mA·cm−2) | 25.12 | 25.27 |
8. Thermal Stability
Long-term stability is the remaining obstacle that prevents perovskite solar cells from displacing established photovoltaic technologies in the market. I therefore subjected unencapsulated devices to continuous thermal aging at 65 °C in a nitrogen atmosphere and tracked the normalized power conversion efficiency over time. The control device degrades steadily and retains only about 55% of its initial efficiency after 900 h. The 4HPCl-modified device degrades much more slowly and retains approximately 90% of its initial efficiency over the same period.
I attribute this improvement to two factors that operate in tandem. First, the reduction in residual PbI2 and in the density of grain-boundary defects removes the sites at which thermally activated ion migration and decomposition reactions are most likely to nucleate. Second, the chemical bonds formed between 4HPCl and the perovskite lattice, as evidenced by the X-ray photoelectron spectroscopy shifts, increase the activation energy for the detachment of surface species and for the diffusion of iodide. The combination of a cleaner grain-boundary network and a more strongly bound surface is what produces the observed stability enhancement. The stability comparison is summarized in Table 8.
| Aging time at 65 °C in N2 (h) | Control retained PCE (%) | 4HPCl-modified retained PCE (%) |
|---|---|---|
| 0 | 100 | 100 |
| 900 | ≈ 55 | ≈ 90 |
9. Mechanistic Synthesis
The body of evidence that I have assembled points to a coherent multidimensional passivation mechanism. I can summarize the causal chain as follows.
In the precursor solution, 4HPCl forms soluble complexes with the lead iodide species. These complexes modify the supersaturation profile and delay nucleation, which, through the inverse dependence of grain size on nucleation rate, produces larger grains with a reduced grain-boundary area. During crystallization, the chloride counterion participates in the halide coordination environment and suppresses the formation of residual PbI2, while the organic cation is progressively excluded from the bulk lattice and accumulates at grain boundaries and at the film surface. At these locations the pyridine nitrogen, the hydrazine group, and the protonated nitrogen site engage undercoordinated Pb2+ and iodine vacancies through coordination, hydrogen bonding, and electrostatic interaction, as confirmed by the core-level shifts in the X-ray photoelectron spectra.
The consequences of this passivation are expressed at three levels. At the material level, the trap density falls from 4.68 × 1015 cm−3 to 3.02 × 1015 cm−3, the carrier lifetime rises from 403.77 ns to 1056.12 ns, and the photoluminescence quantum yield more than doubles. At the device level, the leakage current falls by a factor of six, the recombination resistance doubles, the built-in potential rises from 0.66 V to 0.87 V, and the hysteresis index falls from 6.75% to 1.93%. At the system level, the power conversion efficiency rises from 22.38% to 25.31%, and the thermal stability improves from roughly 55% retention to roughly 90% retention after 900 h. I find it particularly satisfying that the material-level, device-level, and system-level metrics all point in the same direction, because this internal consistency gives me confidence that the mechanism I have proposed is the actual cause of the improvement rather than a post hoc rationalization.
The dependence of performance on additive concentration deserves a final comment. If passivation were a simple saturation process, I would expect the performance to increase monotonically with concentration up to a plateau. Instead, I observe an optimum at 0.50 mg·mL−1, with degradation at 1.00 mg·mL−1. At the higher concentration the additive apparently interferes with crystal growth, as evidenced by the broader diffraction peaks and by the reduced photovoltaic performance. This suggests that there is a narrow window in which the additive is abundant enough to passivate the majority of accessible defect sites but not so abundant that it disrupts the crystallization pathway. Identifying this window empirically, as I have done here, remains an essential step in the development of any additive-based passivation strategy for perovskite solar cells.
10. Conclusion
I have demonstrated that the hydrochloride salt 4HPCl functions as a multifunctional passivator for inverted perovskite solar cells when introduced directly into the precursor ink. The molecule combines a pyridine nitrogen, an electron-rich hydrazine group, and a chloride counterion, and each of these motifs contributes a distinct passivation channel. The pyridine nitrogen coordinates to undercoordinated lead, the hydrazine group participates in hydrogen bonding and coordination with the iodide sublattice, and the chloride anion compensates halide vacancies and helps to reconstruct the local coordination environment. The result is a simultaneous reduction in lead-related and halide-related deep-level defects.
The optimized device, fabricated with an additive concentration of 0.50 mg·mL−1, achieves a champion power conversion efficiency of 25.31%, an open-circuit voltage of 1.166 V, and a fill factor of 83.40%, compared with 22.38%, 1.105 V, and 78.06% for the control. The efficiency gain is supported at the material level by an increase in the average carrier lifetime from 403.77 ns to 1056.12 ns and a more than twofold increase in the photoluminescence quantum yield, and at the device level by a reduction in the trap density from 4.68 × 1015 cm−3 to 3.02 × 1015 cm−3, a reduction in the leakage current by a factor of six, and an increase in the built-in potential from 0.66 V to 0.87 V. The devices also exhibit excellent thermal stability, retaining approximately 90% of their initial efficiency after 900 h at 65 °C, compared with approximately 55% for the control.
Beyond the specific numbers, I believe the broader significance of this work lies in the demonstration that a single small molecule can address multiple defect populations simultaneously when its functional groups are chosen to match the distinct chemical environments present at grain boundaries and surfaces. The rational design of such multifunctional passivators, guided by the mechanistic picture that I have developed here, offers a practical and generalizable route toward perovskite solar cells that are simultaneously more efficient and more durable.