Silent convergence failures in SCAPS-1D: unphysical open-circuit voltages and unflagged anomalies in temperature-dependent Perovskite Solar Cell simulations

Abstract Perovskite solar cells (PSCs) based on inorganic charge transport layers are attracting increasing attention for their thermal stability and low fabrication cost, yet the temperature-dependent behavior of carrier extraction in these devices remains incompletely understood. Here we show that numerical simulations using SCAPS-1D can fail silently: temperature sweeps of CuI/MAPbI3/ETL and Spiro-OMeTAD/MAPbI3/ETL architectures, performed across three electron transport layers (SnO2, TiO2, ZnO) and two absorber thicknesses (100 nm and 500 nm) over the range 250–350 K, reveal two distinct convergence failure modes, neither of which is reliably flagged by the solver. In one, open-circuit voltage (Voc) and fill factor (FF) become unresolvable while a seemingly ordinary short-circuit current density (Jsc) is still returned. In the other - observed exclusively in Spiro-OMeTAD-based configurations - the solver returns a Voc that exceeds the thermodynamic ceiling Eg/q by up to 2.9×, with no indication that the output is physically impossible. Three candidate mechanisms for the first failure mode - Debye length confinement breakdown, conduction band offset inversion, and extrinsic-to-intrinsic carrier transition - were analytically ruled out. We demonstrate instead that it is caused by thermally activated divergence of the series resistance Rs at the ETL/perovskite interface: Rs increases by two to four orders of magnitude at anomaly onset, following Arrhenius behavior (R² = 0.88–0.99) with activation energies of 0.59–0.87 eV consistent with thermionic emission across the conduction band spike ΔEc. Both failure modes are ETL/HTL-specific and perovskite thickness-independent - identical patterns at 100 nm and 500 nm confirm an interfacial rather than bulk origin. While this study covers a limited set of architectures, the central implication is general: convergence anomalies in temperature-dependent SCAPS-1D simulations do not always announce themselves, and physically invalid or incomplete results may have gone unnoticed in prior work. These results provide a physically grounded framework for identifying such failures and establish ETL conduction band alignment and HTL identity as key design parameters governing extraction stability and Voc validity, respectively.