Simulated and Real: Tin Perovskite Solar Cell Hits 10.76% After Model-Guided Redesign

Tin-based perovskite solar cells have long promised to deliver lead-free photovoltaics without sacrificing performance, but their fragile chemistry has kept them trailing their lead-based rivals. Now a research team has linked computer prediction and laboratory reality directly: by systematically optimizing every layer of a cesium tin iodide (CsSnI3) solar cell with the SCAPS-1D device simulator and then fabricating the optimized architecture in the lab, the researchers pushed their experimental device to a champion power conversion efficiency of 10.76 percent, up from 8.98 percent for an unoptimized counterpart built with the same materials and environment. The work demonstrates that careful parameter selection guided by numerical modelling can translate into measurable gains in real devices, even when the theoretical ceiling remains far out of reach.

The story begins with the absorber itself. CsSnI3 is a narrow-band-gap tin halide perovskite with a band gap of roughly 1.31 eV, which allows it to absorb a wider swath of the visible spectrum than most lead-based or germanium-based perovskites. That property makes it especially attractive for tandem solar cells, where a narrow-gap top cell paired with a silicon bottom cell can exceed the single-junction Shockley-Queisser limit. But tin carries a notorious weakness: Sn2+ oxidizes to the more stable Sn4+ state when exposed to air, breaking down the perovskite structure. Since the first CsSnI3 solar cell was reported in 2012 with an efficiency of about 0.9 percent, researchers have fought oxidation with additives, encapsulation, and interface engineering, gradually lifting efficiencies past 10 percent in inverted and interface-engineered devices.

In the new study, the team simulated an n-i-p architecture of FTO/SnO2/CsSnI3/Spiro-OMeTAD/Au using SCAPS-1D, a one-dimensional drift-diffusion simulator developed at Ghent University that solves Poisson’s equation and the carrier continuity equations under illumination. The simulation ran at 300 K under AM 1.5G light of 1000 W/m2, with all defects modelled as neutral Gaussian traps of characteristic energy 0.1 eV and capture cross-sections of 10−15 cm2. The researchers then swept one parameter at a time in a fixed sequence: absorber thickness, absorber defect density, Spiro-OMeTAD acceptor concentration and thickness, and SnO2 donor concentration and thickness. The authors acknowledge that this sequential approach cannot capture interactions between parameters, so a full multidimensional global sensitivity analysis was beyond the scope of the work.

The absorber thickness sweep revealed a clear saturation point. As the CsSnI3 layer thickened from 0.01 nm toward 1000 nm, the simulated efficiency climbed rapidly as the cell absorbed more photons and generated more carriers, boosting the short-circuit current density. But around 500 nm the gains flattened because light absorption was essentially complete, while carriers generated farther from the selective interfaces had longer distances to travel before collection, increasing transport and recombination losses. The fill factor initially rose slightly with thickness, then dipped before settling at a relatively constant value. The team identified 500 nm as the optimum absorber thickness within the investigated range.

Defect density proved to be the absorber’s most punishing parameter. When the team raised the CsSnI3 defect density from 1013 to 1018 cm−3, performance held steady up to 1014 cm−3, but from 1015 cm−3 onward the numbers collapsed. At the maximum density of 1018 cm−3, the simulated efficiency and fill factor had plunged to 3.09 percent and 39.42 percent, with the short-circuit current falling to 12.40 mA/cm2 and the open-circuit voltage dropping at a high, constant rate to 0.63 V. The mechanism is classic Shockley-Read-Hall recombination: more traps shorten the diffusion length, which according to the relation LD = Dτ decreases with carrier lifetime, so carriers recombine before they can be collected. The open-circuit voltage, being highly sensitive to trap density, degraded first and fastest. The optimized defect density was set at 1014 cm−3.

The transport layers told a more nuanced story. In the Spiro-OMeTAD hole-transport layer, raising the acceptor concentration improved the open-circuit voltage, an effect the researchers explain using the assumption given by Koster and colleagues, while the short-circuit current deteriorated slightly, which they attribute to impurity scattering centres slowing hole transfer. Performance saturated at an acceptor density of 1.38 × 1018 cm−3, and remarkably, the simulated results were almost independent of Spiro-OMeTAD thickness, likely because heavy doping reduces the layer’s resistance enough to make its dimension nearly irrelevant. The SnO2 electron-transport layer was far more thickness-sensitive. Donor concentration saturated around 2.6 × 1016 cm−3, yielding a simulated efficiency of 27.71 percent, but pushing SnO2 beyond 80 nm degraded efficiency rapidly, bottoming out at 14.03 percent at 400 nm as the increased transport distance hampered carrier collection. Roughly 80 nm was identified as the optimum SnO2 thickness.

Combining all optimized values produced a simulated device with a theoretical efficiency of 28.84 percent, up from 27.17 percent for the unoptimized model, driven by a roughly 10 percent jump in short-circuit current density to 33.31 mA/cm2 while the open-circuit voltage and fill factor slipped about 1 percent to 1.05 V and 82 percent. The simulated external quantum efficiency reached 99 percent across the 400 to 700 nm visible range, with an absorption onset at 950 nm. The energy band diagram showed a positive conduction band offset of 0.27 eV at the SnO2-CsSnI3 interface and a positive valence band offset of 0.309 eV at the absorber-Spiro-OMeTAD interface. The latter exceeds the 0.2 eV threshold above which valence band offsets theoretically prevent good performance, yet the model still delivered excellent outputs, suggesting the barrier is not as prohibitive as the guideline implies in this configuration.

Then came the experimental test. The team fabricated devices using the optimized thickness parameters, dissolving CsI and SnI2 in anhydrous DMF:DMSO with SnF2 added as a reducing additive to suppress Sn2+ oxidation, spin-coating the precursor inside a nitrogen glovebox, and depositing doped Spiro-OMeTAD followed by a thermally evaporated 100 nm gold electrode. X-ray diffraction confirmed the signature orthorhombic CsSnI3 phase with no impurity peaks of Cs2SnI6, and x-ray photoelectron spectroscopy of the deconvoluted Sn-3d peaks, positioned at 494.01 and 485.60 eV, detected no Sn2+ oxidation, indicating the black orthorhombic phase remained intact. The ultraviolet-visible absorption onset at 938 nm corresponded to a band gap of about 1.32 eV, matching the simulation input. Field-emission electron microscopy, however, revealed a rough, nanowire-like film riddled with nanoscale pinholes, which the researchers identify as the primary source of performance loss in CsSnI3 devices.

The fabricated optimized device achieved a champion efficiency of 10.76 percent with a short-circuit current density of 18.90 mA/cm2, an open-circuit voltage of 0.73 V, and a fill factor of 78 percent, compared with 8.98 percent, 17.60 mA/cm2, 0.74 V, and 69 percent for the unoptimized control. The fill factor improvement was the pivotal gain, attributed to better absorber film quality at the optimized thickness, and across 15 to 20 identically fabricated optimized cells the average efficiency was 9.85 percent. The experimental quantum efficiency showed a similar absorption pattern to the simulation up to 950 nm but reached only 55 to 65 percent in the visible region, and the integrated current density of 18.13 mA/cm2 deviated just 4.1 percent from the J-V measurement, confirming internal consistency.

The chasm between the simulated 28.84 percent and the measured 10.76 percent is the study’s most honest finding. SCAPS-1D relies on spatially averaged material and interface parameters and cannot explicitly replicate fabrication-induced contact losses, pinholes, spatially variable defect distributions, local thickness fluctuations, nonuniform composition, or three-dimensional film morphology. For CsSnI3, where Sn2+ oxidation, defect formation, and film morphology dominate device behavior, these limitations bite especially hard. The researchers point to bulk and interfacial defect states, charge transport losses, series and shunt resistances, and energy-level misalignment as the practical factors standing between the idealized model and the laboratory device. Further optimization of film quality, defect passivation, transport-layer interfaces, and fabrication conditions will be necessary to approach the theoretical performance, but the study establishes a direct, quantified link between computational parameter optimization and experimental device development for lead-free tin perovskite photovoltaics.

Subject of Research: SCAPS-1D-guided optimization and experimental fabrication of a lead-free CsSnI3 perovskite solar cell

Article Title: SCAPS-1D-guided optimization and experimental realization of a CsSnI₃-based perovskite solar cell

Article References: Dehingia, A., & Gogoi, H. P. (2026). SCAPS-1D-guided optimization and experimental realization of a CsSnI₃-based perovskite solar cell. Results in Optics, 25, Article 101193. https://doi.org/10.1016/j.rio.2026.101193

Image Credits: AI Generated

DOI: 10.1016/j.rio.2026.101193

Keywords: perovskite solar cell, CsSnI3, lead-free photovoltaics, SCAPS-1D, tin oxidation, device simulation, SnO2 electron transport layer, Spiro-OMeTAD, power conversion efficiency, tandem solar cells, defect density, thin-film solar cells