Simulation and performance analysis of nanostructure-based perovskite solar cells for improving their efficiency Authors: Amir Reza Firouzabadi, Mohammad Reza Shayesteh, Fakhralsadat Rastegari Published: 2026 ABSTRACT Perovskite solar cells (PSCs) offer high efficiency potential; however, their optical absorption is limited by the thin absorber layers required for efficient charge transport. Enhancing light harvesting without degrading electrical performance remains a key challenge for high-efficiency PSCs.In this study, four perovskite-based device architectures are systematically designed and numerically investigated to improve absorption through plasmonic light trapping, interfacial engineering, and tandem integration. Optical properties are analyzed using finite-difference time-domain (FDTD) simulations, while electrical performance is evaluated via coupled finite-element drift–diffusion modeling under standard AM1.5G illumination.The studied configurations include a reference planar structure, a plasmonic design with hemispherical Ag nanostructures, a graphene–InP Schottky-based hybrid architecture, and a perovskite–InP tandem device. The plasmonic structure enhances long-wavelength absorption, increasing power conversion efficiency from 16.1% to 18.9% despite reduced absorber thickness. Incorporation of a graphene–InP interface further improves charge extraction, achieving an efficiency of approximately 24.8%. The tandem configuration provides the broadest spectral utilization, delivering a simulated efficiency of up to 32% through complementary absorption and voltage addition.These results demonstrate the importance of combined optical–electrical optimization and provide design guidelines for developing high-efficiency thin-film perovskite solar cells. Introduction Perovskite solar cells (PSCs) have emerged over the past decade as one of the most transformative photovoltaic technologies, driven by their exceptional optoelectronic properties and unprecedented rate of efficiency improvement. Owing to their tunable bandgap, high absorption coefficient, long carrier diffusion lengths, and low exciton binding energy, hybrid halide perovskites such as CH₃NH₃PbX₃ have enabled power conversion efficiencies exceeding 25%, positioning PSCs among the leading candidates for next-generation thin-film photovoltaics1–3. Beyond solar energy conversion, perovskite materials have also demonstrated remarkable performance in light-emitting diodes, lasers, and photodetectors, underscoring their broad optoelectronic relevance4. The intrinsic material advantages of perovskites—including low Urbach energy, minimal Stokes shift, and high carrier mobility—make them particularly suitable for thin-film architectures5. However, efficient charge extraction typically requires relatively thin absorber layers, generally below ~ 500 nm. In practical solution-processed devices, especially those fabricated via spin-coating or sol–gel routes, maintaining high crystalline quality and suppressing defect-assisted recombination often constrains the perovskite thickness to the range of ~ 280–350 nm, and in certain optimized configurations even below 200 nm 6,7. When the optical penetration depth of incident sunlight exceeds the absorber thickness, incomplete photon absorption occurs, leading to reduced short-circuit current density (Jsc) and limiting overall device performance8,9. This trade-off between optical absorption and carrier collection represents a fundamental bottleneck in thin-film PSC design. In parallel with optical constraints, long-term operational stability remains a major challenge for PSC deployment. Thermal stress, moisture sensitivity, ion migration, and defect formation at grain boundaries and interfaces contribute to nonradiative recombination and performance degradation10. Considerable efforts have therefore focused on compositional engineering, interface passivation, and transport-layer optimization to simultaneously enhance efficiency and stability 11–15. Despite these advances, further performance improvement increasingly depends on advanced light-management strategies capable of enhancing absorption without increasing absorber thickness. Optical and nanophotonic engineering has consequently emerged as a powerful pathway to overcome absorption limitations in thin perovskite films8. Metallic nanostructures—particularly gold and silver—support localized surface plasmon resonances (LSPRs) and surface plasmon polaritons (SPPs), generating strong near-field enhancement and increased optical path length within the active layer16,17. Among various geometries, anisotropic nanostructures such as metal nanorods exhibit stronger and spectrally tunable field enhancement compared to spherical nanoparticles due to their longitudinal plasmon modes6. These plasmonic effects effectively increase the optical path length and absorption cross-section without increasing material thickness. More recently, the integration of atomically thin materials such as graphene has introduced additional opportunities for broadband absorption control and efficient charge extraction, owing to graphene’s high transparency, tunable chemical potential, and low resistive losses18,19.However, direct contact between metallic nanoparticles and the perovskite absorber may induce exciton quenching and interfacial recombination losses. To mitigate these parasitic effects while preserving optical enhancement, core–shell configurations such as Au, SiO₂ have been proposed, where the dielectric shell electrically isolates the metal core from the active layer6. Alternatively, purely dielectric nanostructures, including SiO₂ nanoparticles embedded without introducing metallic recombination centers7. Despite substantial progress, most existing studies investigate enhancement strategies—such as plasmonic scattering, dielectric light trapping, interfacial modification, graphene-assisted extraction, or tandem integration—in isolation, often under disparate modeling assumptions or experimental conditions. As a result, a systematic and quantitative comparison of how successive levels of optical and electrical complexity influence device performance remains lacking, particularly for thin-film perovskite architectures. Furthermore, many reports evaluate optical enhancement using standalone electromagnetic simulations without fully coupling the results to carrier transport and recombination dynamics, potentially overestimating realistic photovoltaic gains. In this work, we address these gaps by presenting a unified and systematic numerical investigation of four perovskite-based solar cell architectures, spanning from a conventional flat-reflector reference design to plasmonic, graphene–Schottky, and perovskite–InP tandem configurations. All scenarios are analyzed within a consistent optoelectronic simulation framework that integrates finite-difference time-domain (FDTD) optical modeling with self-consistent drift–diffusion carrier transport calculations under AM1.5G illumination. This integrated methodology enables a fair and reproducible comparison of absorption enhancement, charge extraction efficiency, recombination losses, and overall photovoltaic performance metrics as a function of structural complexity. By elucidating the trade-offs associated with plasmonic, dielectric, and tandem strategies, this study provides a stepwise roadmap for optimizing thin-film perovskite solar cells and bridges the gap between optical enhancement concepts and realistic device-level performance. Materials and methods Device structures and material selection In this study, four different scenarios for perovskite solar cell structure were investigated to optimize optical and electrical performance. Material selection was based on optical and electrical properties, compatibility with Lumerical FDTD Solutions and FEM simulations, and potential to enhance spectral absorption and charge extraction. The primary absorber layer in perovskite cells was methylammonium lead iodide (MAPbI₃) with a direct bandgap of approximately 1.55 eV, capable of absorbing a wide range of visible light. Other layers included ITO as an anti-reflection coating, silver (Ag) as a continuous or hemispherical nanostructured reflector, TiO₂ as a spacer layer, InP semiconductor substrate and graphene monolayer for Schottky junction formation and charge transport. Optical and electrical properties, including the complex refractive index (n, k), dielectric constant, extinction coefficient, and conductivity, were obtained from experimental databases and peer-reviewed publications and implemented in simulation software. The key electrical parameters used in the device simulations are summarized in Table 1.20–27. Table 1Electrical and physical parameters used for device simulation of MAPbI₃, TiO₂, InP, ITO, graphene, and Ag layers. Values are obtained from literature and adjusted for consistency with the simulation.ParameterMAPbI3TiO2InPITOGrapheneSilver (Ag)Thickness (nm)350–50050Variable80 nm0.34 nm100 nmEg (eV)1.553.251.343.60–χ (eV)3.94.24.384.54.54.26µn (cm2/Vs)20.00659002015,000–µp (cm2/Vs)20.0061502015,000–ND (cm− 3 )01 × 10161 × 10¹⁵1 × 10180–NA (cm− 3 )00000– 20,21 20,21 22,23 24 25,26 27 Table 1Electrical and physical parameters used for device simulation of MAPbI₃, TiO₂, InP, ITO, graphene, and Ag layers. Values are obtained from literature and adjusted for consistency with the simulation.ParameterMAPbI3TiO2InPITOGrapheneSilver (Ag)Thickness (nm)350–50050Variable80 nm0.34 nm100 nmEg (eV)1.553.251.343.60–χ (eV)3.94.24.384.54.54.26µn (cm2/Vs)20.00659002015,000–µp (cm2/Vs)20.0061502015,000–ND (cm− 3 )01 × 10161 × 10¹⁵1 × 10180–NA (cm− 3 )00000– 20,21 20,21 22,23 24 25,26 27 Electrical and physical parameters used for device simulation of MAPbI₃, TiO₂, InP, ITO, graphene, and Ag layers. Values are obtained from literature and adjusted for consistency with the simulation. Optical modeling and governing equations To analyze the propagation of surface plasmon polaritons (SPPs) and the distribution of electromagnetic fields, Maxwell’s equations in the frequency domain were solved using the Helmholtz equation within Lumerical FDTD Solutions. In all simulations, material dispersion was taken into account by using wavelength-dependent refractive indices \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:n\left(\lambda\:\right)$$\end{document}and extinction coefficients \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:k\left(\lambda\:\right)$$\end{document}obtained from experimental optical datasets and implemented through the material libraries/models in Lumerical. The refractive index values reported in Table 3 at specific wavelengths are provided only as representative reference values and were not assumed to be constant over the operating spectral range. For a one-dimensional approximation, it was assumed that the relative permittivity varies only in the vertical (z) direction, and in the multilayer structures, boundary conditions at the interfaces were applied to solve the coupled even and odd modes. The local photogeneration rate in the absorbing layer was calculated based on the imaginary part of the material permittivity and the local electric field. The optical simulation of these multilayer structures was carried out using Lumerical software, considering materials such as MAPbI₃ perovskite, indium tin oxide (ITO), silver (Ag), titanium dioxide (TiO₂), indium phosphide (InP), and graphene, with the optical and geometrical properties of each layer specified as follows: MAPbI₃ perovskite had a refractive index of 2.5 at 550 nm, and the extinction coefficient (k) was extracted from experimental n, k curves and directly imported into the simulation, with an extinction coefficient of approximately 0.1 and a thickness varying between 350 and 500 nm. ITO had optical constants taken from the Cauchy Optical Library, with a refractive index of 1.95 and an extinction coefficient of 0.02 at 550 nm, a thickness of 80 nm, and negligible absorption. Silver (Ag) was modeled using Johnson & Christy and Drude Model, with a refractive index of 0.14 and an extinction coefficient of 3.98 at 550 nm, a thickness of 100 nm, and an absorption of approximately 5%. TiO₂ had optical constants obtained from the Sellmeier, with a refractive index of 2.49 and an extinction coefficient close to zero at 550 nm, a thickness of 50 nm, and negligible absorption. InP had optical constants from Sellmeier, with a refractive index of 3.44 and an extinction coefficient of 0.01 at 550 nm. Graphene was modeled as a two-dimensional material, using a Drude-based (2D Kubo) optical model. An effective refractive index of approximately 2.7 was assumed, while the extinction coefficient was set to 0.05 according to the Kubo conductivity model. The material exhibits an optical absorption of approximately 2.3% per layer within the visible range, which is typically considered between 550 and 800 nm for reference. All simulations were performed using Lumerical software, with the optical and geometrical parameters of each layer defined according to the specified values to accurately investigate the optical response of the multilayer structures. Table 3Optical and geometrical properties of materials used in lumerical simulations.MaterialOptical modelReference λ n kThicknessAbsorptionNotesMAPbI₃Extracted from experimental n, k and Dataset550 nm2.50.10350–500 nm105 cm⁻¹Perovskite active layer, k from experimental dataITOCauchy Optical Library550 nm1.95< 0.0280 nmVery lowTransparent conductive layerAgJohnson & Christy + Drude Model800 nm0.143.98100 nm~ 5%Metallic layer, high reflectanceTiO₂Sellmeier550 nm2.49≈ 050 nmNegligibleAnti-reflection or transport layerInPSellmeier1200 nm3.440.35VariableNot providedSemiconductor for IR, thickness design-dependentGrapheneDrude (2D, Kubo)550–800 nm2.70.050.34 nm2.3%Single-atom layer, constant absorption 2.3% Table 3Optical and geometrical properties of materials used in lumerical simulations.MaterialOptical modelReference λ n kThicknessAbsorptionNotesMAPbI₃Extracted from experimental n, k and Dataset550 nm2.50.10350–500 nm105 cm⁻¹Perovskite active layer, k from experimental dataITOCauchy Optical Library550 nm1.95< 0.0280 nmVery lowTransparent conductive layerAgJohnson & Christy + Drude Model800 nm0.143.98100 nm~ 5%Metallic layer, high reflectanceTiO₂Sellmeier550 nm2.49≈ 050 nmNegligibleAnti-reflection or transport layerInPSellmeier1200 nm3.440.35VariableNot providedSemiconductor for IR, thickness design-dependentGrapheneDrude (2D, Kubo)550–800 nm2.70.050.34 nm2.3%Single-atom layer, constant absorption 2.3% Optical and geometrical properties of materials used in lumerical simulations. Simulation geometries Four main scenarios were investigated. In Scenario 1, the structure consisted of a MAPbI₃ perovskite layer with a thickness of 500 nm, serving as the primary light-absorbing layer, an 80 nm ITO layer acting as an anti-reflection coating to reduce surface reflection and maximize light transmission in the visible wavelength range, and a 100 nm silver (Ag) back reflector to redirect transmitted photons back into the perovskite layer, enhancing absorption (Fig. 1a). Fig. 1Layered schematics of perovskite solar cells and a perovskite/InP tandem architecture with various light-management and reflector design. Layered schematics of perovskite solar cells and a perovskite/InP tandem architecture with various light-management and reflector design. In Scenario 2, the classical back reflector was replaced by an array of hemispherical Ag nanostructures to enhance light trapping. The hemispherical Ag nanoparticles had a diameter of 150 nm and a lattice period of 250 nm, placed on top of the perovskite layer. A 50 nm TiO₂ spacer layer was inserted between the Ag nanostructures and the perovskite to prevent direct contact, avoiding mechanical and chemical interactions, while also assisting in light scattering. The perovskite thickness was reduced to 350 nm to save material and reduce series resistance without significantly affecting optical absorption (Fig. 1b). In Scenario 3, a dual-layer structure was employed, consisting of a 350 nm MAPbI₃ perovskite layer as the top absorber for visible-light absorption (Eg ≈ 1.55 eV), a 0.34 nm graphene monolayer acting as a Schottky contact and transparent conductive layer with ~ 97% transmission, and a patterned InP substrate underneath serving as the bottom layer to enhance near-infrared absorption (Eg ≈ 1.34 eV). The patterned InP was designed to support guided wave modes, further increasing IR absorption. This configuration combines the high visible-light absorption of perovskite with the fast electrical properties of InP and the high conductivity of graphene (Fig. 1c). In Scenario 4, a tandem (cascade) structure was designed, consisting of a top perovskite cell (Eg ≈ 1.55 eV) to absorb high-energy photons and a bottom InP cell (Eg ≈ 1.34 eV) to absorb lower-energy photons. Charge transport layers and current-matching optimization layers were placed between the two subcells. A perovskite back reflector was installed beneath the InP layer to redirect transmitted photons back into the absorbing layers, and an anti-reflection coating was applied at the top surface to minimize input reflection and maximize light absorption across the tandem structure (Fig. 1d). Solar spectrum and AM1.5G implementation To account for standard solar illumination conditions, the AM1.5G solar spectrum from the ASTM G-173 database was used as the incident photon flux. This spectrum corresponds to solar irradiation at an angle of 48.27° and is commonly used as a reference for evaluating the practical performance of photovoltaic devices. For each wavelength, the calculated external quantum efficiency (EQE) was multiplied by the AM1.5G photon flux to obtain the actual spectral distribution of absorbed power. The overall absorption and reflection for each device scenario were then calculated by averaging the wavelength-dependent values weighted according to the AM1.5G spectrum. This ensures that the optical performance metrics reflect realistic solar illumination conditions. Preliminary results indicated that Scenario 2, despite having a thinner perovskite layer, exhibits optical absorption comparable to Scenario 1, while Scenario 4 achieves the broadest spectral coverage and the lowest overall reflection. FDTD optical simulation All structures were simulated using Lumerical FDTD Solutions. Boundary conditions included perfectly matched layers (PML) in the vertical direction and periodic or PML boundaries in lateral directions. A broadband light source covering 350–900 nm with TM polarization was used for SPP studies. Outputs included absorption \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:A\left(\lambda\:\right)$$\end{document}, reflection \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:R\left(\lambda\:\right)$$\end{document}, transmission \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:T\left(\lambda\:\right)$$\end{document}, near-field intensity \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\mid\:E{\mid\:}^{2}$$\end{document}, and SPP coupling efficiency. FEM electrical simulation Charge transport was modeled using FEM software (Lumerical DEVICE Simulator) by solving the Poisson equation and drift–diffusion equations self-consistently. Recombination mechanisms included radiative, Shockley–Read–Hall, and Auger processes. The coupled equations are:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\nabla\:\cdot\:\left(\epsilon\:\nabla\:\phi\:\right)=-q\left(p-n+{N}_{D}^{+}-{N}_{A}^{-}\right)$$\end{document}2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{J}_{n}=q{\mu\:}_{n}n\nabla\:\phi\:+q{D}_{n}\nabla\:n$$\end{document}3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{J}_{p}=q{\mu\:}_{p}p\nabla\:\phi\:-q{D}_{p}\nabla\:p$$\end{document}4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\frac{\partial\:n}{\partial\:t}=\frac{1}{q}\nabla\:\cdot\:{J}_{n}+G-R$$\end{document}5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\frac{\partial\:p}{\partial\:t}=-\frac{1}{q}\nabla\:\cdot\:{J}_{p}+G-R$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\phi\:\:$$\end{document}is the electrostatic potential, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:n\:$$\end{document}and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:p\:$$\end{document}are electron and hole densities, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mu\:}_{n}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\mu\:}_{p}\:$$\end{document}are mobilities, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{D}_{n}$$\end{document}and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{D}_{p}$$\end{document} are diffusion coefficients, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:G\:$$\end{document}is the optical generation rate, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:R\:$$\end{document}is the total recombination rate. Optical–electrical coupling Optical generation rates obtained from Lumerical FDTD simulations were imported into FEM simulations as the spatially-resolved generation term \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:G(x,y,z,\lambda\:)$$\end{document}. This procedure ensures accurate modeling of charge extraction and current distribution, correlating photon absorption to carrier transport. Device performance analysis Key performance metrics such as current–voltage (J–V) characteristics, power conversion efficiency (PCE), open-circuit voltage (Voc), short-circuit current (Jsc), and fill factor (FF) were calculated using MATLAB. Data visualization, curve fitting, and sensitivity analyses were performed with Python using libraries such as NumPy, SciPy, pandas, matplotlib, and seaborn. Results Spectral optical response (R, T, A) The simulated reflectance (red line), transmittance (blue line), and absorbance (green line) spectra for the four proposed device configurations are presented in Fig. 2. In all scenarios, the transmittance (blue line) remains near zero across the 350–800 nm range, confirming the effectiveness of the Ag back reflectors in preventing optical leakage. Fig. 2Reflectance, transmittance, and absorbance spectral profiles for each scenario. Reflectance, transmittance, and absorbance spectral profiles for each scenario. As shown in Fig. 2a, the baseline structure (Scenario 1) exhibits a notable reduction in reflectance within the mid-visible range (500–650 nm), where the absorbance reaches its peak of approximately 65%. Although the minimum reflectance occurs in this region, the lack of advanced light-trapping mechanisms leads to a relatively high reflectance (above 40%) at shorter and longer wavelengths, limiting the overall photon harvest. In Scenario 2 (Fig. 2b), the integration of 150 nm diameter hemispherical Ag nanoparticles leads to a pronounced reduction in reflectance, specifically forming a deep dip near 700 nm. This corresponds to a significant absorbance enhancement in the NIR region, where the green curve rises to nearly 75%. This improvement results from the localized surface plasmon resonance (LSPR) and far-field scattering induced by the nanostructured back reflector, which effectively increases the optical path length within the reduced 350 nm perovskite layer. Figure 2c illustrates that Scenario 3 achieves a broader and more stable spectral absorption profile. The graphene–InP interface and the periodic InP substrate contribute to reduced reflectance across a wider band, particularly in the 450–550 nm region where reflectance drops to nearly 20%. The absorbance remains consistently above 70% from 500 nm to 800 nm. This demonstrates that the structured InP layer enhances optical coupling by suppressing the interference-driven fluctuations seen in the planar baseline. Scenario 4 (Fig. 2d) demonstrates the most complex optical behavior, featuring two distinct absorption maxima near 500 nm and 700 nm. These peaks correspond to the complementary bandgaps of the perovskite top cell and the InP bottom cell, respectively. The reflectance curve shows a characteristic oscillatory pattern, indicating successful spectral splitting between the subcells. This tandem architecture achieves the broadest spectral coverage, maximizing the utilization of both the visible and NIR portions of the solar spectrum. External quantum efficiency (EQE) The EQE spectrum for the reference structure (Fig. 3a) shows a prominent peak of approximately 87% in the mid-visible region. The EQE gradually decreases toward longer wavelengths due to the intrinsic decline in the perovskite absorption coefficient. The response remains stable in the blue–green spectral region, reflecting adequate refractive index matching provided by the front ITO layer. This configuration is considered the baseline structure for performance comparison. Fig. 3External Quantum Efficiency (EQE) curves of the four proposed design scenarios. External Quantum Efficiency (EQE) spectra of the four investigated design scenarios. Scenario 1 (baseline) exhibits a peak EQE of ≈ 87%, primarily in the visible range. Scenario 2 incorporates Ag hemispheres and TiO₂, enhancing the long-wavelength and near-infrared (NIR) response and increasing the peak EQE to ≈ 90%. Scenario 3 adds a Graphene–InP Schottky junction, further improving performance to a peak EQE of ≈ 92% and providing up to 10% enhancement in the 500–700 nm range. Scenario 4 (Tandem structure) achieves the highest performance, with a maximum EQE of ≈ 95% and the broadest spectral coverage from 300 to 900 nm. These results demonstrate that the proposed structural optimizations progressively enhance both efficiency and spectral response. External Quantum Efficiency (EQE) curves of the four proposed design scenarios. External Quantum Efficiency (EQE) spectra of the four investigated design scenarios. Scenario 1 (baseline) exhibits a peak EQE of ≈ 87%, primarily in the visible range. Scenario 2 incorporates Ag hemispheres and TiO₂, enhancing the long-wavelength and near-infrared (NIR) response and increasing the peak EQE to ≈ 90%. Scenario 3 adds a Graphene–InP Schottky junction, further improving performance to a peak EQE of ≈ 92% and providing up to 10% enhancement in the 500–700 nm range. Scenario 4 (Tandem structure) achieves the highest performance, with a maximum EQE of ≈ 95% and the broadest spectral coverage from 300 to 900 nm. These results demonstrate that the proposed structural optimizations progressively enhance both efficiency and spectral response. As presented in Fig. 3b, Scenario 2 yields a 5–10% increase in EQE within the 600–700 nm band relative to Scenario 1, resulting in a maximum EQE of approximately 90%. Enhanced light trapping induced by plasmonic scattering results in a clear improvement in the NIR region. The shape of the EQE curve in the mid-visible region remains largely comparable to that of the baseline structure, while demonstrating an overall upward shift in the 600–700 nm range. Figure 3c indicates that Scenario 3 achieves an EQE peak approaching 92%, accompanied by a broadening of the effective spectral response. Enhancements of ~ 10% within 500–700 nm are observed, attributable to the combined effects of reduced optical losses at the graphene interface and optical coupling provided by the patterned InP layer. The tandem device (Fig. 3d) exhibits two major EQE maxima around 92–95% corresponding to the visible-absorbing perovskite subcell and the NIR-absorbing InP subcell. Across 300–900 nm, Scenario 4 consistently surpasses the other configurations, demonstrating the most complete spectral utilization. Electrical performance (J–V characteristics) Figure 4a summarizes the electrical output parameters for Scenario 1, where the device consists of an ITO antireflection coating (ARC) combined with an Ag mirror as the back reflector. The short-circuit current density (Jsc) is approximately 15.24 mA/cm², and the open-circuit voltage (Voc) is around 1.30 V. The fill factor (FF) is 0.81, leading to a power conversion efficiency (PCE) of approximately 16.1%. The combination of the ITO ARC and Ag mirror provides a good balance, yielding satisfactory photocurrent generation and minimal resistive losses. Fig. 4Comparison of electrical performance across the four scenarios. Comparison of electrical performance across the four scenarios. In Scenario 2, featuring an Ag hemispherical back reflector, the Jsc improves to approximately 17.12 mA/cm², while the Voc increases slightly to 1.36 V. The FF rises to 0.82, attributed to enhanced light trapping and reduced recombination losses. The resulting PCE is approximately 18.9%, marking a significant improvement compared to Scenario 1. These results highlight the beneficial effects of nanostructured back reflectors in improving photon recycling while maintaining electrical performance. Figure 4c shows that the optimized version of Scenario 3 achieves Jsc ≈ 21.8 mA cm⁻², Voc ≈ 1.36 V, and FF ≈ 0.84, yielding η ≈ 24.8%. The improved charge extraction facilitated by the graphene–InP interface and the reduced series resistance contribute to the superior performance relative to Scenarios 1 and 2. Scenario 4 (Fig. 4d) delivers the strongest overall performance among the investigated configurations, with a short-circuit current density of Jsc ≈ 24.5 mA cm⁻², an open-circuit voltage of Voc ≈ 1.45 V, and a fill factor of FF ≈ 0.93, resulting in a power conversion efficiency of η ≈ 32%. The enhanced performance arises from voltage addition between the perovskite and InP subcells combined with complementary absorption across the visible and near-infrared regions of the solar spectrum. It should be emphasized that the reported efficiency of approximately 32% for the perovskite–InP tandem configuration represents a simulated upper-limit performance under idealized yet physically consistent assumptions. The primary objective of this analysis is to elucidate relative performance trends and design trade-offs rather than to predict immediate experimental realizability. Comparative evaluation of optical and electrical metrics A comprehensive comparison of the four configurations is provided here, incorporating the quantitative and qualitative data previously detailed in Tables 3 and 4 into the following discussion. Scenario 1 serves as a stable baseline with a simple and low-cost structure, achieving an efficiency (η) of ≈ 16.1% and a fill factor (FF) of 0.81. While it provides a reliable reference, it suffers from limited absorption enhancement due to the reliance on a flat Ag mirror and the absence of advanced light-trapping mechanisms.Scenario 2 offers a noticeable improvement, reaching an efficiency of ≈ 18.9% and a higher FF (≈ 0.82) by utilizing Ag hemispherical nanostructures (150 nm diameter). These structures enable efficient light trapping at longer wavelengths (λ > 750 nm), allowing for higher output even with a reduced absorber thickness of 350 nm. However, this improvement comes at the cost of increased fabrication complexity and the need for precise control over nanostructure uniformity and illumination alignment.Scenario 3 leverages a monolayer graphene–InP Schottky interface to achieve superior charge extraction, resulting in a significant jump in efficiency to ≈ 24.8% (Jsc ≈ 21.8 mA/cm2; FF ≈ 0.84). This architecture benefits from interface passivation and improved Schottky contact, though it requires stringent control over graphene doping and faces thermal-compatibility challenges during InP–perovskite integration. Finally, Scenario 4 provides the highest overall PCE of ≈ 32% and a Voc of ≈ 1.45 V. This tandem configuration utilizes full spectral coverage across the visible-NIR regions through a perovskite top cell and an InP bottom cell, despite the high material costs and the extreme complexity of precise spectral alignment and tunnel-junction control. Table 4Strength–weakness assessment of Scenarios 1–4 based on optical and electrical performance metrics.ScenarioStrengthsWeaknesses1Simple and low-cost structure; acceptable Jsc and Voc; stable fill factor (FF ≈ 0.81) with overall efficiency of ≈ 16.1%; suitable baseline reference for evaluating advanced architecturesLimited potential for absorption enhancement using only ARC and a flat Ag mirror; absence of advanced light-trapping mechanisms2Noticeable improvement in Jsc and FF (≈ 0.82), leading to efficiency of ≈ 18.9%; higher output at reduced thickness; efficient light trapping at longer wavelengthsIncreased fabrication complexity and production cost; need for precise control of illumination alignment and uniformity of hemispherical Ag arrays3Enhanced carrier collection due to interface passivation and improved Schottky contact; increased Jsc and FF, resulting in η ≈ 24.8%; strong potential for hybrid integration of III–V materials with perovskiteRequirement for careful control of graphene doping and Schottky barrier properties; complex growth processes and thermal-compatibility challenges in InP–perovskite integration4Highest Voc (due to voltage addition of subcells); improved Jsc and FF leading to total efficiency of ≈ 32%; full spectral coverage across visible–NIR and optimized current matchingHighly complex fabrication; need for precise spectral alignment and strict control of the tunnel-junction; high material and epitaxial growth costs Table 4Strength–weakness assessment of Scenarios 1–4 based on optical and electrical performance metrics.ScenarioStrengthsWeaknesses1Simple and low-cost structure; acceptable Jsc and Voc; stable fill factor (FF ≈ 0.81) with overall efficiency of ≈ 16.1%; suitable baseline reference for evaluating advanced architecturesLimited potential for absorption enhancement using only ARC and a flat Ag mirror; absence of advanced light-trapping mechanisms2Noticeable improvement in Jsc and FF (≈ 0.82), leading to efficiency of ≈ 18.9%; higher output at reduced thickness; efficient light trapping at longer wavelengthsIncreased fabrication complexity and production cost; need for precise control of illumination alignment and uniformity of hemispherical Ag arrays3Enhanced carrier collection due to interface passivation and improved Schottky contact; increased Jsc and FF, resulting in η ≈ 24.8%; strong potential for hybrid integration of III–V materials with perovskiteRequirement for careful control of graphene doping and Schottky barrier properties; complex growth processes and thermal-compatibility challenges in InP–perovskite integration4Highest Voc (due to voltage addition of subcells); improved Jsc and FF leading to total efficiency of ≈ 32%; full spectral coverage across visible–NIR and optimized current matchingHighly complex fabrication; need for precise spectral alignment and strict control of the tunnel-junction; high material and epitaxial growth costs Strength–weakness assessment of Scenarios 1–4 based on optical and electrical performance metrics. The evolution of Jsc, Voc, FF, and η across the scenarios, as illustrated in Table 6 and Fig. 5, confirms that performance consistently improves from Scenario 1 to Scenario 4, with Scenario 4 achieving the most balanced optical–electrical synergy (Tables 5 and 6). Table 5Key optical features, electrical performance implications, and design complexity of the proposed device architectures.ScenarioOptical focusKey optical featuresAbsorber thickness (nm)Electrical focusPredicted impactComplexity/costNotes1Reduction of reflectance across 300–800 nmITO layer (~ 80 nm) serving as ARC + Ag back reflector for photon recycling500↑ Jsc due to enhanced absorption; Voc characteristic of single-junction devices↑ absorption in visible region; moderate improvement at longer wavelengthsLow to mediumBaseline, stable, and simple design; serves as the reference configuration2Enhanced light-trapping for λ > 750 nmAg hemispherical nanostructures (diameter ≈ 150 nm, pitch ≈ 250 nm) + TiO₂ spacer (~ 50 nm)350Maintains or increases Jsc at reduced absorber thickness; reduced recombination↑ absorption in NIR; PCE ≈ 18.9% (optimized single-junction performance)Medium (requires nanopatterning)Plasmonic scattering and light trapping improve efficiency without increasing thickness3Balanced absorption in visible and NIR; reduced optical losses at contactMonolayer graphene as Schottky contact + periodic InP substrate–↑ carrier extraction; FF ≈ 0.84; Jsc ≈ 21.8 mA cm⁻²; η ≈ 24.8%Improved carrier extraction and reduced ohmic losses; superior to single-layer structuresMedium to high (graphene + InP integration)Innovative architecture requiring band-alignment tuning and optimized contact formation4Full utilization of visible + NIR spectrumPerovskite top cell + InP bottom cell with ARC and metallic reflector–Voc ≈ 1.45 V; Jsc ≈ 24.5 mA cm⁻²; η ≈ 32%Broadest EQE coverage; dual improvement in Voc and Jsc; highest theoretical efficiencyHigh (tandem fabrication + current matching demands)Highest optical–electrical potential; requires precise spectral alignment and complex fabrication Table 5Key optical features, electrical performance implications, and design complexity of the proposed device architectures.ScenarioOptical focusKey optical featuresAbsorber thickness (nm)Electrical focusPredicted impactComplexity/costNotes1Reduction of reflectance across 300–800 nmITO layer (~ 80 nm) serving as ARC + Ag back reflector for photon recycling500↑ Jsc due to enhanced absorption; Voc characteristic of single-junction devices↑ absorption in visible region; moderate improvement at longer wavelengthsLow to mediumBaseline, stable, and simple design; serves as the reference configuration2Enhanced light-trapping for λ > 750 nmAg hemispherical nanostructures (diameter ≈ 150 nm, pitch ≈ 250 nm) + TiO₂ spacer (~ 50 nm)350Maintains or increases Jsc at reduced absorber thickness; reduced recombination↑ absorption in NIR; PCE ≈ 18.9% (optimized single-junction performance)Medium (requires nanopatterning)Plasmonic scattering and light trapping improve efficiency without increasing thickness3Balanced absorption in visible and NIR; reduced optical losses at contactMonolayer graphene as Schottky contact + periodic InP substrate–↑ carrier extraction; FF ≈ 0.84; Jsc ≈ 21.8 mA cm⁻²; η ≈ 24.8%Improved carrier extraction and reduced ohmic losses; superior to single-layer structuresMedium to high (graphene + InP integration)Innovative architecture requiring band-alignment tuning and optimized contact formation4Full utilization of visible + NIR spectrumPerovskite top cell + InP bottom cell with ARC and metallic reflector–Voc ≈ 1.45 V; Jsc ≈ 24.5 mA cm⁻²; η ≈ 32%Broadest EQE coverage; dual improvement in Voc and Jsc; highest theoretical efficiencyHigh (tandem fabrication + current matching demands)Highest optical–electrical potential; requires precise spectral alignment and complex fabrication Key optical features, electrical performance implications, and design complexity of the proposed device architectures. Table 6Comparison of the key parameters (η, FF, Voc, and Jsc) for four scenarios.ScenarioJsc (mA/cm²)Voc (V)FFη (%)115.241.300.8116.1217.121.300.8218.9321.81.360.8424.8424.51.450.9332 Table 6Comparison of the key parameters (η, FF, Voc, and Jsc) for four scenarios.ScenarioJsc (mA/cm²)Voc (V)FFη (%)115.241.300.8116.1217.121.300.8218.9321.81.360.8424.8424.51.450.9332 Comparison of the key parameters (η, FF, Voc, and Jsc) for four scenarios. Fig. 5Visual evaluation of optical and electrical attributes across the four scenarios. Visual evaluation of optical and electrical attributes across the four scenarios. Discussion The systematic comparison of the four investigated device architectures provides clear insight into how successive levels of optical and interfacial complexity influence the performance of thin-film perovskite solar cells (Table 5). By evaluating all configurations within a unified optical–electrical simulation framework, the present study isolates the individual and combined contributions of light management, charge extraction, and spectral utilization to key photovoltaic metrics, including EQE, Jsc, Voc, FF, and overall power conversion efficiency. This coupled framework avoids overestimating purely optical gains by explicitly accounting for recombination and resistive losses in the electrical model. For the reference structure (Scenario 1), the combination of a front 80 nm ITO antireflection coating (ARC) and a flat Ag back reflector effectively suppresses surface reflection and enhances absorption within the perovskite absorber. This configuration provides a stable baseline; however, planar ARCs typically minimize reflection only at specific wavelengths rather than across the broad solar spectrum. As a result, while optical impedance matching is improved at targeted wavelengths, broadband spectral utilization remains sub-optimal6. Nonetheless, this scenario demonstrates that careful optical design, including optimization of layer thickness and refractive index, can achieve respectable current densities and efficiencies in thin perovskite films, consistent with previous reports28,29. Introducing plasmonic nanostructures in Scenario 2 leads to a pronounced improvement in optical absorption, particularly at longer wavelengths where thin perovskite layers typically suffer from incomplete absorption. This enhancement arises from the excitation of localized surface plasmon (SP) resonances, with the nanostructures acting as subwavelength “antennas” that efficiently couple light into the active layer6. Hemispherical Ag nanostructures, separated from the perovskite by a TiO₂ spacer, improve performance through two complementary mechanisms: (i) far-field scattering that increases the effective optical path length via angular redistribution of incident photons photons—a phenomenon strongly dependent on nanostructure diameter and spacing7, and (ii) ) near-field plasmonic enhancement, which locally amplifies the electromagnetic field intensity. These regions of intense field localization, commonly referred to as plasmonic hotspots, are primarily concentrated at the Ag/TiO₂ and TiO₂/perovskite interfaces, driving the observed increase in the photogeneration rate. The combined effect of these far-field and near-field interactions results in a spectrally broadened absorption enhancement rather than a narrow resonance peak. Consequently, these improvements translate into higher EQE and Jsc values while maintaining a high fill factor, demonstrating that plasmonic back reflectors can compensate for reduced absorber thickness without significant electrical penalties. The TiO2 dielectric spacer plays a crucial role by electrically isolating the metal nanostructures from the perovskite, thereby preventing exciton quenching and direct charge recombination at the metallic surface6. However, the simulations also indicate that excessive nanostructure density or volume may increase recombination and parasitic absorption, underscoring the need for careful geometric optimization of plasmonic elements30–32. This trade-off emphasizes that optical enhancement must always be balanced against additional metallic losses and interface-induced recombination pathways. Scenario 3 explores the impact of interfacial engineering through the integration of a graphene–InP Schottky junction beneath the perovskite absorber. The results show that this architecture can outperform purely plasmonic designs by improving carrier extraction and reducing series resistance, leading to a higher Jsc and overall efficiency. This correlates with Experimental Impedance Spectroscopy (EIS) data which shows that optimized interfacial nanostructures significantly reduce series resistance (Rs) while increasing recombination resistance (Rrec)6. From a band-alignment perspective, the graphene–InP interface facilitates selective carrier extraction while suppressing back-carrier diffusion, provided that the Schottky barrier height is properly engineered. While optical coupling and broadband absorption are enhanced through the InP substrate and graphene interface, the electrical performance remains sensitive to interfacial losses and energy-level alignment. This highlights that, although graphene-based Schottky contacts offer attractive transparency and conductivity, their full potential depends on precise interface control, including graphene doping, surface passivation, and contact engineering to minimize recombination losses30. Additionally, the use of transparent conductive oxides like ITO as an interconnect remains vital to ensure optical transparency and reduce resistive losses7. The tandem configuration (Scenario 4) demonstrates the highest simulated performance, benefiting from complementary absorption in the perovskite and InP subcells and voltage addition across the junction. Rather than implying a violation of fundamental limits, the efficiency enhancement arises from multi-junction operation, which redistributes thermalization and transmission losses across sub-cells with different bandgaps. To achieve such high performance, current matching between the sub-cells is a fundamental requirement, often requiring the optimization of the bottom cell thickness to match the current generated by the top perovskite layer7. The broadened spectral utilization leads to increased photocurrent generation and a higher open-circuit voltage relative to single-junction architectures. As observed in our results, an optimized nanostructured tandem can lead to a significant enhancement in current density (around 8.3% or more) compared to its planar counterpart7. Importantly, these results represent an upper-bound performance scenario derived under idealized yet physically consistent assumptions. Thus, the reported efficiencies should be interpreted as technologically feasible targets under optimized fabrication conditions rather than guaranteed experimental values. The reduction in hysteresis and the high FF observed are consistent with the improved electrical properties and charge collection efficiency provided by nanostructured interfaces6. In practice, tandem device efficiency will be influenced by current-matching constraints, interlayer resistances, and recombination at the tunnel or recombination junctions. Consequently, the primary value of this configuration lies in illustrating the theoretical performance ceiling and guiding future optimization strategies rather than predicting immediate experimental outcomes33,34. A key outcome of this work is the demonstration that optical enhancement alone does not guarantee proportional electrical performance gains. Only through coupled optical–electrical modeling can the balance between absorption enhancement, carrier transport, and recombination be accurately assessed. The present framework emphasizes the importance of reproducing simulation conditions—such as material dispersion models, boundary conditions, and illumination parameters—during experimental implementation to ensure that predicted optical benefits translate into measurable electrical improvements29. From a practical perspective, the comparative analysis suggests that Scenario 2 offers the most favorable balance between performance enhancement and fabrication complexity. This is particularly true given that such nanostructured layers can be deposited using cost-effective, simple spin-coating techniques rather than expensive conventional lithography7. This relative fabrication simplicity strengthens the technological relevance of plasmonic back reflectors compared to more complex tandem or Schottky-based architectures. While advanced architectures such as graphene–Schottky and tandem devices exhibit higher theoretical efficiencies, their realization requires additional processing steps and stringent interface control. Plasmonic back reflectors based on Ag nanostructures, by contrast, can be integrated using established nanofabrication techniques and provide substantial gains in light trapping with relatively low implementation risk. Nevertheless, challenges such as angular sensitivity, spectral selectivity, and long-term stability of metallic nanostructures remain and may be mitigated through protective coatings or alternative plasmonic materials31. Finally, several limitations of the present study should be acknowledged. The simulations assume idealized energy-level alignment, neglect long-term degradation mechanisms, and do not account for thermal or chemical effects such as Ag oxidation or ion migration under prolonged illumination. Moreover, mechanical stress, interdiffusion at interfaces, and large-area uniformity constraints were not explicitly modeled. Future work should extend this framework to include multiphysics modeling—incorporating optical, electrical, and thermal effects—as well as systematic experimental validation to bridge the gap between numerical predictions and device realization. Conclusion This study presents a systematic optical–electrical comparison of four perovskite-based solar cell architectures to clarify the impact of structural and interfacial design on thin-film device performance. Using a unified and reproducible simulation framework, the contributions of light trapping, plasmonic enhancement, charge extraction, and spectral utilization were quantitatively evaluated.The results show that plasmonic back reflectors based on hemispherical silver nanostructures effectively enhance absorption and current generation in thin perovskite layers while maintaining favorable electrical characteristics, offering a practical balance between performance improvement and fabrication complexity. More advanced designs, including graphene–InP Schottky hybrid devices and perovskite–InP tandem cells, deliver higher simulated efficiencies through improved charge extraction and complementary absorption but are more sensitive to interfacial losses and current-matching constraints. Consequently, their performance should be regarded as indicative of upper-bound design trends rather than immediate experimental benchmarks.Overall, the comparative roadmap established here highlights the necessity of coupled optical–electrical modeling for accurately assessing efficiency–complexity trade-offs in perovskite photovoltaics. These findings provide guidance for selecting and optimizing device architectures and support the rational development of scalable, high-efficiency thin-film perovskite solar cells.