Uncertainty-aware computational screening of planar perovskite solar-cell stacks using a validated TMM--drift-diffusion--WKB solver
Year: 2026
Type: article
Abstract: Deterministic device screening can create a misleadingly precise ranking when thin-film lifetime, trap capture, interface recombination, and parasitic resistance are treated as fixed material constants. We therefore separate an optimistic numerical ceiling from evidence-bounded uncertainty in a computational screen of 432 planar perovskite solar-cell stacks. The design space combines two architectures, two transparent electrodes, three electron-transport layers, four absorbers, three hole-transport layers, and three rear contacts. An adaptive architecture-aware voltage-bracketing and validity protocol yields 428 valid solutions and four solved low-collection designs. Under the archived optimistic library, the highest result is n-i-p FTO/TiO 2 /MAPbI 3 /NiO/Au at 25.882% power-conversion efficiency; this value is a ceiling, not an experimental prediction. In a first-stage uncertainty analysis of 12 compositionally and architecturally diverse fixed-geometry designs (32 Latin-hypercube samples per design), the same stack gives 22.451/24.013/24.541% at the 5th/50th/95th percentiles and leads 30 of 32 paired samples within the selected set. A separate FAPbI 3 explicit-trap diagnostic uses photovoltaic-film trap density, electron and hole capture coefficients, trap energy, two interface velocities, and area-normalized series and shunt resistance. Its leading TiO 2 /FAPbI 3 /NiO design gives 11.883/19.373/22.812%, with bulk trap density the dominant sensitivity (median absolute Spearman correlation 0.855). These calculations support conditional hypothesis prioritization, not experimental validation of a global ranking. A separate literature-constrained mobile-ion Stage-1 pilot screens scan-rate-dependent dynamic hysteresis for 12 fixed geometries: 258/288 cases pass the declared numerical gates, and the evidence-eligible n-i-p FTO/TiO 2 /MAPbBr 3 /NiO/Au design has worst-rate normalized HI 1.33 × 10 −5 ∕3.28 × 10 −5 ∕6.00 × 10 −5 at p05/p50/p95. This is a conditional risk screen and does not predict electrochemical degradation, stabilized maximum-power operation, or operational stability.
Source: Next Energy
Authors: Ahmed Nabih, Ahmed H. Zaki, Osama Helal
Institution: Military Technical College
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $1,280
OpenAlex ID: W7214014331