View in EDS

Design Optimization of Graphite Retort and Predictive Modeling of Optical Properties in Multi-crystalline Silicon for Solar Cell Applications.

Saved in:
Bibliographic Details
Title: Design Optimization of Graphite Retort and Predictive Modeling of Optical Properties in Multi-crystalline Silicon for Solar Cell Applications.
Authors: Raman, Mariyappan1,2 (AUTHOR), Arun, S.3 (AUTHOR), Srinivasan, M.1,2 (AUTHOR) srinisastri@gmail.com, Thamotharan, Keerthivasan4 (AUTHOR), Sree, R. M. Kaviya5 (AUTHOR), Krishnamurthy, Vallidevi5 (AUTHOR), Usami, Noritaka4 (AUTHOR)
Source: Journal of Electronic Materials. Jun2026, Vol. 55 Issue 6, p5321-5340. 20p.
Subjects: Directional solidification, Machine learning, Thermal stresses, Silicon crystals, Reflectance, Solar cell design
Abstract: This work employs numerical simulation of the directional solidification (DS) process to investigate strategies for enhancing multi-crystalline silicon growth. The study examines the effects of modifying the graphite retort's dimensions on impurity concentration, power consumption, melt–crystal interface properties, thermal stress, and dislocation density. Melt–crystal interface changes were analyzed at 25%, 50%, and 75% of the solidification phase. Among the tested conditions, multi-crystalline silicon ingots grown under the optimized case 3 configuration (retort width 25 mm) exhibited superior quality, with dislocation density reduced by above 106 cm−2, impurity concentration decreased by 1014–1015 atoms/cm2, and reflectivity prediction accuracy of 12.76146%. Although case 3 incurred a slight increase in power consumption, the quality improvements make it suitable for semiconductor and solar cell applications. Machine learning (ML) analysis was applied as a tool to better monitor and control silicon quality rather than as the primary objective. To support this, a multi-crystalline silicon ingot was experimentally grown under case 3 conditions and sliced into wafers. Random forest demonstrated excellent performance in predicting minimal reflectivity, while XGBoost efficiently estimated average and maximum reflectivity, as well as minority carrier lifetime improvements, key parameters for industrial solar cell applications. Unlike previous studies focusing only on process optimization, this work uniquely integrates retort geometry optimization, experimental validation, and ML-based reflectivity prediction, providing a comprehensive framework for high-quality silicon ingot production. [ABSTRACT FROM AUTHOR]
Copyright of Journal of Electronic Materials is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:This work employs numerical simulation of the directional solidification (DS) process to investigate strategies for enhancing multi-crystalline silicon growth. The study examines the effects of modifying the graphite retort's dimensions on impurity concentration, power consumption, melt–crystal interface properties, thermal stress, and dislocation density. Melt–crystal interface changes were analyzed at 25%, 50%, and 75% of the solidification phase. Among the tested conditions, multi-crystalline silicon ingots grown under the optimized case 3 configuration (retort width 25 mm) exhibited superior quality, with dislocation density reduced by above 106 cm−2, impurity concentration decreased by 1014–1015 atoms/cm2, and reflectivity prediction accuracy of 12.76146%. Although case 3 incurred a slight increase in power consumption, the quality improvements make it suitable for semiconductor and solar cell applications. Machine learning (ML) analysis was applied as a tool to better monitor and control silicon quality rather than as the primary objective. To support this, a multi-crystalline silicon ingot was experimentally grown under case 3 conditions and sliced into wafers. Random forest demonstrated excellent performance in predicting minimal reflectivity, while XGBoost efficiently estimated average and maximum reflectivity, as well as minority carrier lifetime improvements, key parameters for industrial solar cell applications. Unlike previous studies focusing only on process optimization, this work uniquely integrates retort geometry optimization, experimental validation, and ML-based reflectivity prediction, providing a comprehensive framework for high-quality silicon ingot production. [ABSTRACT FROM AUTHOR]
ISSN:03615235
DOI:10.1007/s11664-026-12843-9