Moisture-induced degradation pathways in mixed-cation lead halide perovskites: an in situ environmental TEM study

Abstract

Moisture remains a central degradation stress for mixed-cation lead halide perovskites, but the earliest nanoscale pathways are difficult to separate from beam damage, thermal artefacts, and post-mortem rearrangement. We report a low-dose in situ environmental transmission electron microscopy study of FA0.83MA0.12Cs0.05Pb(I0.83Br0.17)3 thin films exposed to controlled water vapour at room temperature. Plan-view lamellae on silicon nitride windows were imaged under pulsed-beam conditions while selected-area electron diffraction, energy-dispersive X-ray spectroscopy, photoluminescence mapping, grazing-incidence X-ray scattering, and time-of-flight secondary-ion mass spectrometry were used as correlative checks. Three degradation pathways were resolved. First, water adsorbed at undercoordinated grain-boundary and surface facets, producing reversible lattice swelling and diffuse hydrate-like contrast within 90-180 s at 0.9 mbar H2O. Second, prolonged exposure generated iodide-rich and bromide-rich nanoscale domains, with halide redistribution beginning at grain boundaries and triple junctions before appearing inside grains. Third, irreversible PbI2 platelet formation and Pb-rich residues developed after local cation loss, especially in regions with pre-existing PbI2 inclusions or high misorientation. Raising water pressure from 0.3 to 2.4 mbar reduced the median time to irreversible PbI2 nucleation from 18.6 to 5.2 min. Cs-containing regions delayed hydrate contrast but did not prevent halide redistribution. Beam-blanked and dry-gas controls showed that water vapour, light, and electron dose act synergistically: continuous illumination under water accelerated PbI2 nucleation by a factor of 2.1, while continuous electron irradiation alone produced Pb-rich features that were morphologically distinct from water-driven platelets. The results provide a phase-resolved degradation sequence for mixed-cation perovskites and show why moisture-stability claims require coupled environmental, optical, and low-dose electron-microscopy controls.

Introduction

Metal halide perovskites combine strong optical absorption, long carrier diffusion lengths, and solution-processable fabrication, but their deployment depends on stability under heat, light, bias, oxygen, and moisture [1,2]. Mixed-cation and mixed-halide compositions improved efficiency and operational robustness relative to early methylammonium lead iodide materials [3,4,5]. Even so, water remains a persistent stress because it can coordinate to organic cations, promote hydrate formation, accelerate ion migration, and trigger conversion to PbI2-rich products [10,11,12,13,14].

Moisture degradation is not a single reaction. In methylammonium lead iodide, reversible hydration, partly reversible phase transitions, hydrolysis, and irreversible decomposition have all been reported depending on humidity, illumination, film morphology, and measurement history [10,11,12,13]. In mixed-cation and mixed-halide films, additional degrees of freedom appear: FA, MA, Cs, iodide, and bromide can redistribute at different rates, and local composition can determine whether water exposure is temporarily accommodated or becomes destructive [15,16,17]. Ion migration and photo-induced halide redistribution further complicate the picture [6,7,8,9].

Environmental transmission electron microscopy offers a way to watch nanoscale changes during gas exposure, but halide perovskites are exceptionally beam sensitive. In situ TEM under bias, heat, light, or gas can reveal real processes, yet it can also create ion migration and decomposition artefacts [18,19,20,21,22,23,24]. A credible ETEM study therefore needs dry controls, beam-blanked controls, dose-rate variation, and correlative non-electron-probe measurements.

This article reports a low-dose in situ ETEM investigation of moisture-induced degradation in a triple-cation mixed-halide perovskite. The goal is not to replace device-scale ageing tests. Instead, we identify where degradation starts, which intermediate features appear before irreversible PbI2 formation, and how water exposure couples to light and electron dose. Throughout, we distinguish direct TEM observations from interpretations supported by diffraction, spectroscopy, and ex situ chemical mapping.

Materials and sample preparation

The investigated absorber composition was FA0.83MA0.12Cs0.05Pb(I0.83Br0.17)3, selected because related mixed-cation and mixed-halide perovskites are representative of high-performance device stacks [3,4,5]. Films were spin-coated on plasma-cleaned silicon nitride TEM windows using an antisolvent route and annealed at 100 deg C for 45 min. The average film thickness was 440 +/- 30 nm on companion glass substrates and 390 +/- 40 nm on TEM windows. Grain size from plan-view SEM was 260 +/- 110 nm.

Two controls were prepared. The first was MAPbI3, included because its moisture chemistry and reversible hydration are well documented [10,11,12,13]. The second was a Cs-rich variant, FA0.75MA0.05Cs0.20Pb(I0.83Br0.17)3, used to test whether additional inorganic cation content changed the early ETEM response. All films were stored in dry nitrogen and transferred to the microscope in a sealed holder to minimise uncontrolled laboratory exposure.

Focused-ion-beam cross-sections were used only for post-mortem comparison. The operando ETEM data came from as-grown plan-view films on electron-transparent windows, avoiding gallium implantation and lamella-sidewall artefacts. Companion samples were deposited on glass/ITO/SnO2 stacks for photoluminescence and GIWAXS measurements under matched humidity. The ETEM window geometry is not a complete solar cell, but it isolates absorber degradation without metal-electrode or transport-layer reactions.

Before water exposure, films were screened by low-dose diffraction and EDX. Regions with obvious pre-existing PbI2 platelets or pinholes were excluded from the primary statistics but retained for a separate defect-sensitivity analysis. This avoids letting one poorly prepared area dominate the degradation sequence.

Environmental TEM protocol

ETEM was performed at 200 kV using differential pumping and a water-vapour leak valve. Water partial pressures were 0.3, 0.9, and 2.4 mbar at room temperature. These values are below ambient relative-humidity exposures but high enough to drive measurable changes in the microscope. A residual-gas analyser confirmed that water was the dominant added species. Dry nitrogen and high-vacuum sequences were recorded before and after each water-exposure run.

Beam dose was controlled by pulsed illumination. The standard imaging condition used a dose rate of 0.35 e A^-2 s^-1 and a cumulative dose below 18 e A^-2 for a complete 30 min movie. Diffraction patterns were collected with short exposures and a selected-area aperture to avoid continuous irradiation of the same region. Additional high-dose runs were performed intentionally to identify beam-driven decomposition. Beam sensitivity and mitigation strategies in hybrid perovskites motivated this protocol [20,21,22].

A subset of experiments included 1 sun-equivalent 530 nm illumination through an optical port. Light exposure was synchronised with beam blanking so that water-plus-light effects could be separated from water-plus-electron effects. Recent on-chip light-incorporated TEM work shows why this separation matters for metal-halide perovskites [23]. We did not apply electrical bias in the ETEM holder; bias-driven degradation has been studied separately and would add another transport pathway [19].

Image stacks were drift-corrected using carbon support markers outside the perovskite film. Regions of interest were segmented into grain interiors, grain boundaries, triple junctions, pre-existing PbI2-rich inclusions, and voids. Diffraction rings and spots were indexed against perovskite, PbI2, and hydrate-like reference spacings, but we avoid over-assigning weak diffuse features because water-containing intermediates can be poorly crystalline.

Correlative measurements

Grazing-incidence wide-angle X-ray scattering was performed on companion films exposed to the same water pressures in a sealed humidity cell. GIWAXS provided ensemble phase information that ETEM lacks. Photoluminescence maps tracked changes in emission peak position and intensity, while ToF-SIMS after exposure mapped iodide, bromide, cesium, and organic-cation fragments. These measurements were not used to claim nanoscale time resolution; they served as checks on ETEM interpretation.

The correlative data were essential for halide segregation. EDX in TEM is sensitive to beam damage and has limited sensitivity for light elements and volatile organic fragments. ToF-SIMS and PL showed that bromide-rich regions persisted after water exposure in areas where ETEM showed boundary-initiated contrast changes. This supports the interpretation that some ETEM contrast reflected real halide redistribution rather than only thickness or diffraction-condition changes.

Device-scale relevance was checked with simple glass/ITO/SnO2/perovskite/spiro-OMeTAD/Au devices exposed to 45% relative humidity under open-circuit conditions. These devices were not optimised for record efficiency. Their purpose was to verify that the same moisture sequence correlated with PL loss and performance decay. After 2 h of exposure, average power-conversion efficiency fell by 18% for the baseline mixed-cation composition and 31% for MAPbI3.

Correlative measurements also helped distinguish reversible from irreversible changes. Some lattice-swelling contrast disappeared after returning to dry nitrogen, while PbI2 platelets and Pb-rich residues remained. GIWAXS showed partial recovery of the perovskite peak after short exposure but not after long exposure at 2.4 mbar. This matches earlier reports of partially reversible hydrate formation and irreversible hydrolysis [10,12,13].

Early hydration at boundaries

The first visible response to water was weak contrast at grain boundaries and triple junctions. At 0.9 mbar H2O, the median onset time was 126 s for the mixed-cation film and 74 s for MAPbI3. The contrast appeared as a narrow bright-dark boundary fringe in phase-contrast imaging and as diffuse scattering near the perovskite reflections. The feature was reversible after brief exposure: returning to dry nitrogen within 5 min removed most boundary contrast and restored PL intensity to within 93% of its initial value.

We assign this stage to water adsorption and hydrate-like lattice swelling rather than to PbI2 formation. The evidence is threefold. First, PbI2 reflections were absent at early times. Second, the perovskite diffraction spots shifted reversibly by 0.13-0.21% in d-spacing. Third, companion GIWAXS showed a broad low-angle shoulder consistent with a water-containing intermediate. This behaviour is consistent with reversible hydration studies of methylammonium lead iodide and partially reversible PbI2-CH3NH3I-H2O phase transitions [10,12].

Boundary onset was heterogeneous. Boundaries with high local misorientation, measured from diffraction-spot splitting, degraded earlier than low-angle boundaries. Pre-existing PbI2 inclusions also acted as initiation sites. This does not mean that PbI2 causes all degradation; rather, PbI2-rich or undercoordinated regions appear to provide adsorption and ion-transport pathways for water. The observation is compatible with broader stability reviews that link morphology and interfacial chemistry to moisture response [1,14].

The Cs-rich variant delayed the median boundary-contrast onset to 212 s at 0.9 mbar. The delay was measurable but not protective indefinitely. After 20 min, boundary contrast and diffuse scattering were present in all compositions. Additional inorganic cation content therefore improved early moisture tolerance but did not remove the boundary pathway.

Halide and cation redistribution

After the hydration stage, boundary contrast spread into adjacent grain interiors and small domains with different diffraction contrast appeared. EDX maps collected after beam blanking showed iodide-rich and bromide-rich patches at length scales of 20-80 nm. ToF-SIMS on companion films confirmed lateral halide redistribution after water exposure. The redistribution began at boundaries and triple junctions before becoming visible inside grains.

Mixed-halide perovskites are known to undergo photo-induced halide redistribution and trap formation under illumination [8,9]. Our data show that water accelerates a related segregation pathway even under weak electron illumination. Under water plus 1 sun-equivalent light, the median time to detectable halide-domain contrast fell by 44% relative to water in the dark. Under dry light exposure, the change was slower and largely reversible. Water therefore changes the kinetic pathway rather than simply reproducing ordinary light-induced segregation.

Cation redistribution was harder to observe directly in ETEM. Cs maps from EDX were noisy but showed slight enrichment in regions that resisted early boundary swelling. SIMS showed depletion of MA-related fragments in the most degraded regions and weaker depletion of FA fragments. This is consistent with the expectation that volatile or weakly bound organic components participate in water-driven decomposition, but the microscopy does not directly measure molecular products.

The interaction between water and organic cations is supported by molecular and spectroscopic studies showing hydrogen bonding and orientational changes in MAPbI3 under water exposure [25]. We therefore interpret the early contrast as a coupled lattice-water-cation response. We do not claim to see individual water molecules or organic-cation rotation in ETEM.

Irreversible PbI2 formation

Irreversible degradation was marked by the appearance of PbI2 platelets and Pb-rich residues. PbI2 nucleation usually occurred at grain boundaries, triple junctions, or pre-existing inclusions. At 0.9 mbar H2O, the median time to first irreversible PbI2 was 11.4 min for the baseline mixed-cation film, 7.1 min for MAPbI3, and 16.8 min for the Cs-rich variant. Increasing water pressure to 2.4 mbar reduced the baseline median to 5.2 min.

The PbI2 platelets had clear lattice fringes and diffraction spacings consistent with 2H-PbI2. Their growth was anisotropic, extending along grain boundaries and exposed facets before thickening. EDX showed local iodine enrichment during early platelet growth, followed by Pb-rich residues after prolonged electron exposure. This sequence is important: water-driven PbI2 platelets and beam-driven metallic Pb features were morphologically distinct in low-dose controls.

Continuous electron irradiation accelerated residue formation. In high-dose water experiments, Pb-rich particles appeared before well-defined PbI2 platelets, resembling known beam-driven halide perovskite decomposition and ion migration [21,22]. In pulsed-beam experiments, PbI2 platelets preceded Pb-rich residues. We therefore treat metallic Pb-like features as late-stage or beam-assisted products, not as the primary water-degradation marker.

Light and water acted synergistically. Under water plus light, PbI2 nucleation was 2.1 times faster than under water in the dark at the same pressure. This agrees with in situ phase-evolution work showing coupled photo- and moisture-dependent degradation [15]. It also explains why device stability under illumination can be worse than dark humidity storage even when the nominal water dose is similar [18].

Beam and gas controls

Dry high-vacuum and dry-nitrogen controls showed slow beam-driven damage but no boundary-first hydrate contrast. At the standard pulsed dose, the perovskite diffraction intensity declined by 6 +/- 2% over 30 min in dry nitrogen. At ten times the dose rate, diffraction intensity fell by 38 +/- 6%, and Pb-rich nanoparticles appeared. This dose sensitivity is consistent with prior beam-damage reports and with pulsed-beam mitigation studies [20,21].

Beam-blanked water exposures provided the most important control. Regions exposed to water for 10 min while the beam was blanked showed boundary contrast and occasional PbI2 platelets when imaged afterward at low dose. The onset times could not be measured during blanking, but the final morphology matched pulsed-beam movies rather than high-dose movies. This demonstrates that water alone can drive the observed boundary pathway.

Water-vapour purity was also checked. Introducing dry oxygen at matched pressure did not reproduce the hydration contrast over the same time window, although oxygen can contribute to degradation under illumination through iodide-defect pathways [13]. The present experiment isolates water as the dominant gas-phase trigger. In real devices, water and oxygen will often act together, and the pathway may differ.

The quasi-two-dimensional Ruddlesden-Popper control from recent literature warns that laser, water vapour, and electron beams can produce synergistic degradation [24]. Our 3D mixed-cation films show the same qualitative lesson: environmental microscopy must be interpreted as a coupled perturbation experiment, not as passive observation.

Quantitative pathway map

Combining all movies, we constructed a pathway map with three event times: boundary hydration onset, halide-domain contrast onset, and irreversible PbI2 nucleation. At 0.9 mbar H2O, the baseline mixed-cation film showed median times of 126 s, 6.3 min, and 11.4 min. MAPbI3 showed 74 s, 4.1 min, and 7.1 min. The Cs-rich variant showed 212 s, 8.9 min, and 16.8 min. The ordering was consistent across more than 120 analysed grains per composition.

Event times scaled approximately with water pressure as t proportional to p_H2O^{-0.7} over the tested range. The exponent should not be over-interpreted because ETEM pressure, surface adsorption, and local film thickness are coupled. It does indicate that the process is not a simple one-molecule surface reaction with linear pressure dependence. Boundary diffusion, local dissolution, and ion redistribution likely all contribute.

Distance-to-boundary statistics confirmed that irreversible PbI2 was not randomly distributed. More than 70% of first PbI2 nuclei appeared within 35 nm of a grain boundary or triple junction, even though that region represented less than 30% of the projected film area. Grain interiors degraded later unless they contained visible defects or strong diffraction-contrast variations. Local crystal misorientation is known to influence optoelectronic quality in halide perovskites [26], and our data suggest that it also influences moisture vulnerability.

The pathway map provides a useful operational picture. Short water exposure causes partly reversible hydration and PL loss. Longer exposure causes halide and cation redistribution. Still longer exposure produces irreversible PbI2 and Pb-rich residues. Mixed-cation engineering shifts the timescales but does not change the order of events.

Discussion

The observations support a boundary-mediated degradation mechanism for mixed-cation perovskite films. Water first interacts with undercoordinated or strained boundary regions, creating a hydrate-like expanded lattice. This stage can partly recover when water is removed. Continued exposure activates ion redistribution, especially halide segregation, and eventually produces irreversible PbI2. The mechanism reconciles reversible hydration reports with irreversible hydrolysis and device degradation reports [10,11,12,13,14].

Mixed cations improve the early response but do not make the film moisture-proof. Cs-rich regions delay boundary swelling, and multiple-cation compositions are known to improve moisture stability [4,17]. However, the ETEM movies show that water still finds fast pathways through boundaries and defects. Stability therefore depends not only on average composition but also on microstructure, boundary chemistry, residual PbI2, and local misorientation.

The study also clarifies what ETEM can and cannot say. ETEM directly resolves morphology, lattice contrast, diffraction changes, and heavy-element redistribution. It does not directly quantify water content, volatile organic fragments, or all light-element chemistry. The assignments of hydrate-like intermediates and cation loss rely on comparison with GIWAXS, PL, SIMS, and prior molecular and spectroscopy studies [10,15,16,25].

For device design, the results point to boundary passivation and encapsulation as complementary requirements. Bulk compositional engineering delays degradation, but boundary-first water attack means that grain-boundary chemistry and film densification remain critical. Operation-condition studies further show that humidity, light, and electrical load cannot be treated independently [18].

Limitations

The ETEM water pressures and sample geometry are not identical to ambient humidity ageing. A thin film on a silicon nitride window has different heat sinking, mechanical constraint, and gas access than a full encapsulated solar cell. The timescales reported here should therefore be read as relative pathway kinetics, not as field lifetime predictions.

The electron beam is an unavoidable perturbation. We reduced dose, used pulsed illumination, and performed beam-blanked controls, but we cannot make the microscope completely invisible. Some late-stage Pb-rich residues are certainly beam assisted. The paper therefore emphasises early boundary hydration and PbI2 nucleation under controlled dose rather than high-dose end states.

Finally, the composition space is limited. The baseline film, MAPbI3 control, and Cs-rich variant are enough to test mixed-cation effects qualitatively, but not to optimise composition. Additives, two-dimensional capping layers, passivating salts, and transport layers could change the degradation sequence. The method can be applied to those systems, but the specific event times will not transfer directly.

Conclusion

Low-dose environmental TEM combined with correlative optical, X-ray, and SIMS measurements resolves a three-stage moisture-degradation pathway in mixed-cation lead halide perovskites. Water first produces partly reversible boundary hydration, then drives halide and cation redistribution, and finally nucleates irreversible PbI2 platelets and Pb-rich residues. The pathway starts at grain boundaries, triple junctions, and pre-existing PbI2-rich defects rather than uniformly inside grains.

Mixed-cation composition delays but does not eliminate the pathway. Cs-rich regions slow early hydration, while water plus light accelerates irreversible PbI2 nucleation. Beam controls show that electron dose can mimic or accelerate some late-stage products, making low-dose and beam-blanked experiments essential. The central practical lesson is that moisture stability is a local microstructural problem as much as a bulk-composition problem.

Data and code availability

The supplementary archive contains raw ETEM movies, diffraction frames, gas-pressure logs, beam-dose records, EDX maps, GIWAXS and PL data, SIMS depth profiles, segmentation masks, event-time tables, and scripts for all statistics. Full uncompressed movies are listed by checksum, with compressed analysis copies included for reproducibility.