Scholarlink

<p>This work presents a theoretical investigation of the photovoltaic potential of the ternary telluride chalcogenide Rb<sub>2</sub>PdTe<sub>2</sub>. This study employs an integrated framework of machine learning and deep learning techniques to evaluate the material’s suitability as a solar absorber. Phonon dispersion calculations confirmed the dynamic stability of Rb<sub>2</sub>PdTe<sub>2</sub>, showing no imaginary frequencies. Electronic structure analysis using the HSE06 hybrid functional revealed an indirect band gap of 1.21&#xa0;eV, well positioned for efficient solar spectrum utilization. Strong absorption across the visible and ultraviolet regions further highlights its optoelectronic potential. Device performance was examined using SCAPS-1D simulations. From 196 configurations of electron and hole transport layers (ETL/HTL), the architecture ITO/WS<sub>2</sub>/Rb<sub>2</sub>PdTe<sub>2</sub>/CuSbS<sub>2</sub>/Ni emerged as optimal, delivering a power conversion efficiency (PCE) of 15.31%, with <i>V</i><sub>OC</sub> = 0.99&#xa0;V, <i>J</i><sub>SC</sub> = 17.98&#xa0;mA&#xa0;cm<sup>−2</sup>, and FF = 85.90%. Optimal absorber parameters included a thickness of 1.5&#xa0;µm, an acceptor density of 1 × 10<sup>18</sup>&#xa0;cm<sup>−3</sup>, and a defect density of 1 × 10<sup>15</sup>&#xa0;cm<sup>−3</sup>. Stability and reliability were further supported through quantum efficiency, recombination, resistance, and temperature-dependent analyses. To accelerate device optimization, we implemented a combined machine learning and deep learning framework trained on the SCAPS-1D dataset. Five models for each were tested, with the random forest (<i>R</i><sup>2</sup> = 0.996) and deep neural network (<i>R</i><sup>2</sup> = 0.998) showing the best accuracy. Feature importance identified absorber defect density as the dominant factor influencing performance. The model’s predictive capability was validated against simulation: For a given input set (absorber thickness, defect density, and temperature), SCAPS-1D predicted a PCE of 15.31%, while RF and DNN output 14.98%, confirming its reliability. Finally, our results establish Rb<sub>2</sub>PdTe<sub>2</sub> as a stable, efficient, and highly promising absorber material for thin-film solar cells, with machine learning offering a powerful tool for rapid device design and optimization.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Next-generation solar absorber candidate Rb2PdTe2: a study based on DFT, SCAPS-1D, and machine learning approaches

  • 期刊论文

  • 发表日期：

    2025年12月14日

查看全文
  • Md. Tarekuzzaman,
  • Md. Hasin Ishraq,
  • Adnan Sami Sarker,
  • Md. Zahid Hasan,
  • Sayed Sahriar Hasan

摘要

This work presents a theoretical investigation of the photovoltaic potential of the ternary telluride chalcogenide Rb2PdTe2. This study employs an integrated framework of machine learning and deep learning techniques to evaluate the material’s suitability as a solar absorber. Phonon dispersion calculations confirmed the dynamic stability of Rb2PdTe2, showing no imaginary frequencies. Electronic structure analysis using the HSE06 hybrid functional revealed an indirect band gap of 1.21 eV, well positioned for efficient solar spectrum utilization. Strong absorption across the visible and ultraviolet regions further highlights its optoelectronic potential. Device performance was examined using SCAPS-1D simulations. From 196 configurations of electron and hole transport layers (ETL/HTL), the architecture ITO/WS2/Rb2PdTe2/CuSbS2/Ni emerged as optimal, delivering a power conversion efficiency (PCE) of 15.31%, with VOC = 0.99 V, JSC = 17.98 mA cm−2, and FF = 85.90%. Optimal absorber parameters included a thickness of 1.5 µm, an acceptor density of 1 × 1018 cm−3, and a defect density of 1 × 1015 cm−3. Stability and reliability were further supported through quantum efficiency, recombination, resistance, and temperature-dependent analyses. To accelerate device optimization, we implemented a combined machine learning and deep learning framework trained on the SCAPS-1D dataset. Five models for each were tested, with the random forest (R2 = 0.996) and deep neural network (R2 = 0.998) showing the best accuracy. Feature importance identified absorber defect density as the dominant factor influencing performance. The model’s predictive capability was validated against simulation: For a given input set (absorber thickness, defect density, and temperature), SCAPS-1D predicted a PCE of 15.31%, while RF and DNN output 14.98%, confirming its reliability. Finally, our results establish Rb2PdTe2 as a stable, efficient, and highly promising absorber material for thin-film solar cells, with machine learning offering a powerful tool for rapid device design and optimization.