Anomalous 3D nanoscale photoconduction in hybrid perovskite semiconductors revealed by tomographic atomic force microscopy

Hybrid perovskite (HP) semiconductors such as MAPbI3 exhibit exceptional optoelectronic properties, yet the role of grain boundaries (GBs) in carrier transport remains debated, with reports of beneficial, detrimental, or neutral effects. Conventional surface-only measurements are confounded by depth inhomogeneities and topographic artifacts, obscuring intrinsic GB behavior. The study aims to determine the true, depth-resolved impact of GBs on photogenerated carrier transport in polycrystalline MAPbI3 thin films by developing and applying tomographic atomic force microscopy (T-AFM) to obtain fully 3D nanoscale maps of photoconduction. Clarifying GB roles is important for optimizing device architectures (e.g., solar cells, photodetectors) via grain boundary engineering.

Prior works report mixed conclusions about GBs in halide perovskites, including beneficial effects on carrier collection, detrimental recombination, and benign behavior, often based on surface-probe techniques susceptible to artifacts (e.g., pc-AFM, KPFM). Theoretical studies have predicted intrinsically benign GBs and twin boundaries in MAPbI3, while experimental observations of ferroelastic twin boundaries suggest potential benign transport pathways. Depth-dependent electrical property variations at GBs have been observed, and 3D optical tomography methods have probed buried recombination, but a direct fully 3D, nanoscale-resolved electrical transport mapping within perovskite films has been lacking. This study addresses these gaps by minimizing surface artifacts and revealing buried GB networks and transport behavior throughout the film thickness.

- Materials and film preparation: MAPbI3 precursor prepared by dissolving 0.159 g MAI, 0.461 g PbI2, and 0.014 g MACl additive in 0.930 g mixed solvent (DMF/DMSO 5:1 v/v). Films spin-coated at 4500 rpm for 20 s with 100 µL toluene antisolvent dripped at 10 s, followed by annealing at 120 °C for 15 min in DMSO vapor. All steps in a nitrogen glovebox. Films deposited on FTO. - Characterization: XRD (Bruker D8 Discover, Cu Kα, step 0.02°), SEM (Zeiss G500) for morphology and microstructure. - AFM platforms: T-AFM and pc-AFM in ambient with Asylum Research MFP-3D IO-AFM; KPFM under nitrogen overpressure with Asylum Research Cypher (equivalent KPFM behavior also in ambient on MFP-3D IO-AFM). Fresh, vacuum-sealed samples used; total measurement times <2 h. - Illumination: Primarily back illumination by a broadband LED (<1 sun, uniform over field of view); also 650 nm diode laser for oblique top illumination in some KPFM. - pc-AFM settings: ORCA cantilever holder (1 pA resolution, 20 nA upper limit). Low-load mapping with Ti/Ir-coated probes (radius ~25 nm, f0 ~75 kHz, k ~2.8 N/m). Bias range explored: 0.6–1.1 V for analysis (0–0.3 V featureless due to lack of built-in field; >1.2 V damages probe coating; negative bias ≤−0.4 V risks damage). Pixel acquisition ~1 ms; ionic currents considered negligible. - KPFM: Non-contact AC mode (delta height ~5 nm) with Ti/Ir-coated probes. Compared surface potentials pre- and post-polishing. - T-AFM nanomachining: Conductive diamond probes (radius ~100 nm, f0 ~300 kHz, k ~50 N/m), deflection setpoint ~50 nm (~2.5 µN force), contact mode, ~2 Hz scan rate. Sequential scans remove material to polish surface and access buried layers while minimizing artifacts. - Tomogram acquisition: 34 consecutive pc-AFM scans over 6 µm × 6 µm, pixel size ~11.7 nm. Lateral resolution ~25 nm (from GB FWHM); mean z-resolution ~16 nm; film thickness ~560 nm. Non-uniform removal corrected by interpolation to rectilinear voxels with majority of interpolations within nearest-neighbor distances. - Data analysis of Iph–V: Montage of photocurrent maps at biases 0.6–1.1 V collected on T-AFM-polished regions (~100 nm below initial surface). Using drift relation Iph = q Nph µ A V / L; on polished flat surfaces A and L treated as constant, enabling pixel-wise estimation of µ·Nph from slopes of linear I–V fits. Approximately 32,000 distinct I–V spectra per mapped region. Optical model assumes uniform incident photon flux I0, absorption coefficient α ~ 6 × 10^4 cm−1, and grain-orientation-insensitive α, η, τ so that Nph is approximately constant, allowing interpretation of µ·Nph maps as relative mobility maps. - Controls and artifact mitigation: Demonstrated inversion of apparent GB behavior before vs after polishing, elimination of spurious KPFM contrast after polishing, demonstrating suppression of topography-related artifacts. Dark and zero-bias conditions produce featureless maps consistent with absence of built-in field. - Reproducibility: Voltage and tomographic measurements repeated multiple times with equivalent results.

- 3D photoconduction mapping: T-AFM reveals a fully 3D, interconnected network of grain boundaries serving as conducting channels throughout polycrystalline MAPbI3 films. - Depth-resolved inversion of GB response: At the as-received rough surface, GBs show lower photocurrent (~0.3 nA) than grains (~3 nA). After T-AFM polishing ~100 nm below the surface, GB photocurrent increases ~20× to ~6 nA while grain photocurrent remains ~3 nA. Enhanced GB conduction persists down to at least ~430 nm depth. - Surface potential behavior: KPFM shows higher surface potential at GBs on rough as-grown surfaces; after nanoscale polishing, GB/grain potential contrast is eliminated, indicating removal of surface-related artifacts or reconstructions. - Bias-dependent mapping and linearity: pc-AFM maps from 0.6–1.1 V show uniform enhancement of GB photocurrent with bias. Pixel-wise linear fits yield µ·Nph maps largely uniform within grains but enhanced at GBs. About 32,000 I–V spectra analyzed per region. Near 0–0.3 V, maps are featureless due to lack of built-in field (films on FTO). - Relative mobility inference: Assuming approximately constant Nph, µ·Nph maps indicate higher effective carrier mobility at GBs than within grains. - Identification of two GB types: - Type I: Enhanced vertical conduction channels; promote through-thickness transport and may preempt lateral inter-grain diffusion. Many such GBs are buried and only detectable tomographically. - Type II: Insipid/benign interfaces with photocurrent indistinguishable from adjacent grains; lateral transport uninhibited and vertical conduction unenhanced. Estimated up to ~5% of GBs are Type II. - Spatial resolution and film metrics: Lateral resolution ~25 nm; z-resolution ~16 nm; film thickness ~560 nm; uniform illumination; absorption coefficient assumed α ≈ 6 × 10^4 cm−1. - Voltage constraints and robustness: High SNR maps obtained for 0.6–1.1 V; negative or excessive positive biases degrade images or probes. Findings are robust across repeated measurements and independent of scan direction.

The 3D nanoscale mapping demonstrates that grain boundaries in MAPbI3 are not universally detrimental; instead, many (Type I) form highly connected vertical conduction pathways with higher effective carrier mobility than grains. This resolves contradictions in prior surface-only studies by showing that surface roughness, reconstructions, and artifacts can invert apparent GB behavior, whereas buried GBs exhibit intrinsic enhanced photoconduction. The discovery of coexisting benign (Type II) GBs clarifies how some polycrystalline films may display single-crystal-like transport locally. These insights directly inform device design: in lateral architectures, minimizing Type I GBs (and favoring Type II) could reduce scattering and improve diffusion, while vertical devices (solar cells, photodetectors) may benefit from a high fraction of Type I GBs to facilitate through-thickness carrier extraction. The T-AFM approach overcomes the limitations of surface-confined probes, enabling direct visualization of buried transport pathways critical for optimizing HP optoelectronics via GB engineering.

This work introduces tomographic AFM as a means to obtain fully 3D nanoscale photoconduction maps in hybrid perovskite thin films, revealing that: (i) many GBs (Type I) are enhanced conduction channels forming an interconnected network through the film, and (ii) a minority (~5%) are benign (Type II) with grain-like behavior. Depth-resolved measurements reconcile conflicting literature by eliminating surface artifacts and exposing buried GB behavior. The findings provide a framework for grain boundary engineering tailored to device geometry—Type II GBs favored for lateral transport, Type I for vertical transport. Future work should extend T-AFM to fully assembled devices under operating conditions, and combine with atomic-resolution electron microscopy and advanced optical/x-ray probes to unambiguously identify the structural origins (e.g., twins, Ruddlesden–Popper faults) of Type II boundaries and chemically tailor GBs to enhance device performance.

- Device context: Measurements performed on MAPbI3 films on FTO without built-in junction fields; near short-circuit biases (0–0.3 V) maps are featureless. Fully assembled device conditions were not examined. - Voltage range: Analysis restricted to 0.6–1.1 V due to probe/sample limitations; negative and >1.2 V biases degrade measurement fidelity. - Mobility quantification: Only relative mobility (via µ·Nph) inferred; absolute µ requires knowledge of Nph, A, and L with higher precision and assumptions (e.g., constant η, τ, and α across grains). - Surface modification: T-AFM involves controlled material removal; while evidence suggests minimal subsurface damage and improved fidelity, it alters the original surface. - Structural attribution: Type II GBs are hypothesized (e.g., twins or RP faults) but not definitively identified; multimodal correlative studies are needed. - Generalizability: Results pertain to the studied MAPbI3 films and preparation conditions; extrapolation to other compositions/processing requires validation. - Environmental/temporal factors: Measurements limited to short durations (<2 h); long-term stability and environmental effects on 3D transport were not assessed.