Saquib Ahmed
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Saquib Ahmed has a total of 122 co-authors. Here are the top co-authors by number of shared publications:
Tohidul Islam
10 shared publicationsKazi Md. Shorowordi
9 shared publicationsSadiq Shahriyar Nishat
7 shared publications- Sadiq Shahriyar Nishat
Department of Materials Science and Engineering, Rensselaer Polytechnic Institute, Troy, New York 12180, USA
6 shared publications Ahsiur Rahman Nirjhar
Department of Materials and Metallurgical Engineering (MME), Bangladesh University of Engineering and Technology (BUET), East Campus, Dhaka 1000, Bangladesh
5 shared publications- Miah Abdullah Sahriar
Department of Materials and Metallurgical Engineering (MME), Bangladesh University of Engineering and Technology (BUET) 1 , East Campus, Dhaka 1000, Bangladesh
5 shared publications
Latest Publications
- Open Access
- 10 September 2026
- Published by Elsevier in Next Materials
- Vol. 13
Optimization of charge transport layers for lead-free K2AgSbBr6 double perovskite solar cells: A combined DFT and SCAPS-1D study
- 10 September 2026
- Published by Elsevier in Next Materials
- Vol. 13
Open AccessOptimization of charge transport layers for lead-free K2AgSbBr6 double perovskite solar cells: A combined DFT and SCAPS-1D study
- Open Access
- 5 August 2026
- Published by Elsevier in SSRN Electronic Journal
Physics-Constrained Machine Learning Framework for Mechanical Property Prediction,in CNT–Al Composites
The reliable prediction of mechanical properties in carbon nanotube-reinforced aluminum (CNT–Al) composites remains challenging because available experimental datasets are typically sparse, heterogeneous, and incomplete. This study develops a physics-guided machine-learning framework for predicting yield strength, ultimate tensile strength, and hardness from limited CNT–Al composite data. Experimental records collected from published studies were systematically parsed, standardized, and processed using a physics-constrained imputation strategy designed to preserve established directional relationships between material descriptors and mechanical properties. Following data preprocessing, 195 usable samples comprising nine input features and three target variables were retained for model development. Eleven supervised regression algorithms were evaluated using five-fold cross-validation, among which the Extra Trees regressor achieved the highest performance on the literature-derived dataset, with an average R2 of 0.9556 and an RMSE of 21.740. To mitigate the limitations associated with the small sample size, a physics-informed synthetic-data generation scheme was subsequently employed to produce 10,000 samples while maintaining prescribed material bounds and feature–target trends. Among the six models evaluated on the synthetic dataset, the multilayer perceptron produced the best overall performance, with an average R2 of 0.9584 and an RMSE of 7.4074. Model-level sensitivity analysis further indicated that the predicted responses followed the expected monotonic trends associated with key microstructural and processing parameters. The framework was also validated against in-house experimental data for Ni-coated multi-walled CNT reinforced aluminum composite, with a pooled mean absolute percentage error of 10.8% for the three target properties. The results demonstrate the power of physics-based constraints, data augmentation and machine learning to enable mechanical-property prediction in sparse materials datasets, and provide a transferable framework for data-limited composite systems.
- 5 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessPhysics-Constrained Machine Learning Framework for Mechanical Property Prediction,in CNT–Al Composites
The reliable prediction of mechanical properties in carbon nanotube-reinforced aluminum (CNT–Al) composites remains challenging because available experimental datasets are typically sparse, heterogeneous, and incomplete. This study develops a physics-guided machine-learning framework for predicting yield strength, ultimate tensile strength, and hardness from limited CNT–Al composite data. Experimental records collected from published studies were systematically parsed, standardized, and processed using a physics-constrained imputation strategy designed to preserve established directional relationships between material descriptors and mechanical properties. Following data preprocessing, 195 usable samples comprising nine input features and three target variables were retained for model development. Eleven supervised regression algorithms were evaluated using five-fold cross-validation, among which the Extra Trees regressor achieved the highest performance on the literature-derived dataset, with an average R2 of 0.9556 and an RMSE of 21.740. To mitigate the limitations associated with the small sample size, a physics-informed synthetic-data generation scheme was subsequently employed to produce 10,000 samples while maintaining prescribed material bounds and feature–target trends. Among the six models evaluated on the synthetic dataset, the multilayer perceptron produced the best overall performance, with an average R2 of 0.9584 and an RMSE of 7.4074. Model-level sensitivity analysis further indicated that the predicted responses followed the expected monotonic trends associated with key microstructural and processing parameters. The framework was also validated against in-house experimental data for Ni-coated multi-walled CNT reinforced aluminum composite, with a pooled mean absolute percentage error of 10.8% for the three target properties. The results demonstrate the power of physics-based constraints, data augmentation and machine learning to enable mechanical-property prediction in sparse materials datasets, and provide a transferable framework for data-limited composite systems.
- Open Access
- 2 August 2026
- Published by Elsevier in SSRN Electronic Journal
Mechanistic and Data-Driven Modelling of Mechanical Properties in High-Entropy Alloys
Machine-learning models are increasingly used to predict the mechanical properties of high-entropy alloys, yet their apparent accuracy can be inflated by repeated measurements of the same or closely related compositions. Here we combine gradient-boosting regression, composition-aware validation and a closed-form strengthening model to examine yield strength and ultimate tensile strength in a curated dataset of 1,659 records. Although conventional random cross-validation produced values of 0.889 for yield strength and 0.957 for ultimate tensile strength, the dataset contained only 390 distinct canonical compositions. Grouping all records of the same composition within a single fold reduced performance to and , respectively, revealing a substantial generalization penalty that is obscured by random data partitioning. Model-attribution analysis identified test temperature, valence-electron-related descriptors, atomic-size mismatch, elemental-volume statistics and grain-size-related terms as the principal predictive factors, in broad agreement with established metallurgical trends. The closed-form model reproduced the characteristic strength scale of representative FCC-like and BCC-like alloys without fitting to the present dataset, providing a mechanistic reference for interpreting the data-driven predictions. Screening of 544 candidate compositions further recovered physically plausible strength regimes across refractory-BCC-like and FCC-like alloy families. These results establish composition-aware validation as a necessary standard for assessing machine-learning models in high-entropy-alloy design and provide a framework for combining mechanistic insight with data-driven prediction.
- 2 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMechanistic and Data-Driven Modelling of Mechanical Properties in High-Entropy Alloys
Machine-learning models are increasingly used to predict the mechanical properties of high-entropy alloys, yet their apparent accuracy can be inflated by repeated measurements of the same or closely related compositions. Here we combine gradient-boosting regression, composition-aware validation and a closed-form strengthening model to examine yield strength and ultimate tensile strength in a curated dataset of 1,659 records. Although conventional random cross-validation produced values of 0.889 for yield strength and 0.957 for ultimate tensile strength, the dataset contained only 390 distinct canonical compositions. Grouping all records of the same composition within a single fold reduced performance to and , respectively, revealing a substantial generalization penalty that is obscured by random data partitioning. Model-attribution analysis identified test temperature, valence-electron-related descriptors, atomic-size mismatch, elemental-volume statistics and grain-size-related terms as the principal predictive factors, in broad agreement with established metallurgical trends. The closed-form model reproduced the characteristic strength scale of representative FCC-like and BCC-like alloys without fitting to the present dataset, providing a mechanistic reference for interpreting the data-driven predictions. Screening of 544 candidate compositions further recovered physically plausible strength regimes across refractory-BCC-like and FCC-like alloy families. These results establish composition-aware validation as a necessary standard for assessing machine-learning models in high-entropy-alloy design and provide a framework for combining mechanistic insight with data-driven prediction.
- 2 June 2026
- Published by Elsevier in Solar Energy
- Vol. 315
Machine learning assisted development of lead-free Cs3Bi2I9 perovskite solar cells
- 2 June 2026
- Published by Elsevier in Solar Energy
- Vol. 315
Machine learning assisted development of lead-free Cs3Bi2I9 perovskite solar cells
- 9 April 2026
- Published by Science Publishing Group in American Journal of Energy Engineering
- Vol. 14 (2) , 71-82
Perovskite and Cadmium Telluride in Next-generation Photovoltaic Technologies
This study provides a comprehensive comparative analysis of cadmium telluride (CdTe) and perovskite materials as a top-layer for next-generation tandem photovoltaic (PV) technologies, benchmarked against crystalline silicon (c-Si). The evaluation integrates three primary domains, scientific and engineering performance, economic feasibility, and ethical, legal and social implications (ELSI) to establish a multidimensional framework for sustainable material selection. Scientific and engineering assessment analyzed device efficiency, material stability, and degradation behavior. Economic viability was assessed using levelized cost of energy (LCOE) and price per watt (PPW). The ELSI evaluation focused on environmental toxicity, health risks and sourcing concerns such as the use of conflict-region materials. A qualitative assessment of each category was translated to a quantitative scale to enable direct comparison. Results indicate that CdTe exhibits higher stability and lower degradation, maintaining superior durability and manufacturing compared to perovskite. However, perovskites rapidly increasing efficiency, inexpensive raw materials, and declining production costs highlight its long-term promise. Despite CdTe’s strong scientific and ELSI performance, its reliance on tellurium limits scalability, white perovskites environmental challenges, particularly lead toxicity, require further technological innovation. Ongoing research into encapsulation and defect passivation may significantly improve perovskite’s stability and sustainability. Additional, integration of artificial intelligence and machine learning in material discovery can accelerate optimization across all domains. This study underscores the potential for both CdTe and perovskite to transform next-generation solar power systems by combining technical, economic and ethical perspectives. The findings serve as a framework for future research in sustainable PV development, bridging laboratory innovation with real-world energy transition strategies.
- 9 April 2026
- Published by Science Publishing Group in American Journal of Energy Engineering
- Vol. 14 (2) , 71-82
Perovskite and Cadmium Telluride in Next-generation Photovoltaic Technologies
This study provides a comprehensive comparative analysis of cadmium telluride (CdTe) and perovskite materials as a top-layer for next-generation tandem photovoltaic (PV) technologies, benchmarked against crystalline silicon (c-Si). The evaluation integrates three primary domains, scientific and engineering performance, economic feasibility, and ethical, legal and social implications (ELSI) to establish a multidimensional framework for sustainable material selection. Scientific and engineering assessment analyzed device efficiency, material stability, and degradation behavior. Economic viability was assessed using levelized cost of energy (LCOE) and price per watt (PPW). The ELSI evaluation focused on environmental toxicity, health risks and sourcing concerns such as the use of conflict-region materials. A qualitative assessment of each category was translated to a quantitative scale to enable direct comparison. Results indicate that CdTe exhibits higher stability and lower degradation, maintaining superior durability and manufacturing compared to perovskite. However, perovskites rapidly increasing efficiency, inexpensive raw materials, and declining production costs highlight its long-term promise. Despite CdTe’s strong scientific and ELSI performance, its reliance on tellurium limits scalability, white perovskites environmental challenges, particularly lead toxicity, require further technological innovation. Ongoing research into encapsulation and defect passivation may significantly improve perovskite’s stability and sustainability. Additional, integration of artificial intelligence and machine learning in material discovery can accelerate optimization across all domains. This study underscores the potential for both CdTe and perovskite to transform next-generation solar power systems by combining technical, economic and ethical perspectives. The findings serve as a framework for future research in sustainable PV development, bridging laboratory innovation with real-world energy transition strategies.
- 15 January 2026
- Published by American Chemical Society (ACS) in ACS Applied Nano Materials
- Vol. 9 (4) , 2025-2036
Photoluminescence of CVD Grown WS 2 Monolayers Treated with H 2 SO 4 Vapor: Implications for Photoelectric Conversion Materials
- 15 January 2026
- Published by American Chemical Society (ACS) in ACS Applied Nano Materials
- Vol. 9 (4) , 2025-2036
Photoluminescence of CVD Grown WS 2 Monolayers Treated with H 2 SO 4 Vapor: Implications for Photoelectric Conversion Materials
- 22 April 2025
- Published by Elsevier in Journal of Rare Earths
- Vol. 44 (1) , 137-150
Experimental and first principles study on opto-structural property of hydrothermally synthesized Cu1‒Nd O NPs with enhanced photocatalytic activity
- 22 April 2025
- Published by Elsevier in Journal of Rare Earths
- Vol. 44 (1) , 137-150
Experimental and first principles study on opto-structural property of hydrothermally synthesized Cu1‒Nd O NPs with enhanced photocatalytic activity
- 25 February 2025
- Published by American Chemical Society (ACS) in ACS Applied Electronic Materials
- Vol. 7 (5) , 2012-2021
Magnetic Proximity-Induced Colossal Valley Splitting in WTe2 for Room Temperature Valleytronics
- 25 February 2025
- Published by American Chemical Society (ACS) in ACS Applied Electronic Materials
- Vol. 7 (5) , 2012-2021
Magnetic Proximity-Induced Colossal Valley Splitting in WTe2 for Room Temperature Valleytronics