Predicting band gaps of MOFs on small data by deep transfer learning with data augmentation strategies

Zhang, Zhihui; Zhang, Chengwei; Zhang, Yutao; Deng, Shengwei; Yang, Yun‐Fang; Su, An; She, Yuanbin

doi:10.1039/d3ra02142d

Cited by 10 publications

(8 citation statements)

References 56 publications

(53 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This approach can help to overcome data limitations by utilizing pre-existing knowledge and models. 149,150…”

Section: Structural Designmentioning

confidence: 99%

“…This can help to enhance the diversity and coverage of the data. 149 (ii) Transfer learning. Knowledge and models trained on related open-framework materials (e.g.…”

Section: Structural Design Enhanced By Machine-learningmentioning

confidence: 99%

“…This approach can help to overcome data limitations by utilizing pre-existing knowledge and models. 149,150 (iii) Active learning. By intelligently selecting a subset of samples for experimental characterization and validations based on model predictions, researchers can iteratively update and refine their models.…”

Section: Structural Design Enhanced By Machine-learningmentioning

confidence: 99%

See 2 more Smart Citations

Revolutionizing the structural design and determination of covalent–organic frameworks: principles, methods, and techniques

Liu,

et al. 2024

Chem. Soc. Rev.

Self Cite

View full text Add to dashboard Cite

show abstract

“…This approach can help to overcome data limitations by utilizing pre-existing knowledge and models. 149,150…”

Section: Structural Designmentioning

confidence: 99%

“…This can help to enhance the diversity and coverage of the data. 149 (ii) Transfer learning. Knowledge and models trained on related open-framework materials (e.g.…”

Section: Structural Design Enhanced By Machine-learningmentioning

confidence: 99%

Section: Structural Design Enhanced By Machine-learningmentioning

confidence: 99%

See 1 more Smart Citation

Revolutionizing the structural design and determination of covalent–organic frameworks: principles, methods, and techniques

Liu,

et al. 2024

Chem. Soc. Rev.

Self Cite

View full text Add to dashboard Cite

show abstract

“…The hybrid quantum-MOF database (QMOF, 20,375 MOFs), , which consists of experimental and hypothetical computation-ready MOFs collected from previously mentioned databases, has been recently introduced to offer a diverse collection of structurally optimized structures with a comprehensive range of chemical and physical properties . This database was studied for predicting several properties of MOFs such as band gaps , and heat capacities, but it has not been screened for a gas separation application to the best of our knowledge. Motivated by this, we aimed to screen this diverse material space for VOC capture and unlock the potential of QMOFs for adsorption-based C 4 H 10 separation from air for the first time in the literature.…”

Section: Introductionmentioning

confidence: 99%

“…The same group followed a similar approach using GCMC simulations to assess the potential of 31,399 hMOFs in capturing C 3 −C 4 alkanes from an air mixture mimicking VOC/N 2 /O 2 and calculated propane (C 3 H 8 ) selectivities between 10 −3 and 7.3 × 10 5 and butane (C 4 H 10 ) selectivities between 0.1 and 5.7 × 10 6 at 1 bar and 298 K. 32 The hybrid quantum-MOF database (QMOF, 20,375 MOFs), 33,34 which consists of experimental and hypothetical computation-ready MOFs collected from previously mentioned databases, has been recently introduced to offer a diverse collection of structurally optimized structures with a comprehensive range of chemical and physical properties. 22 This database was studied for predicting several properties of MOFs such as band gaps 33,35 and heat capacities, 36 but it has not been screened for a gas separation application to the best of our knowledge. Motivated by this, we aimed to screen this diverse material space for VOC capture and unlock the potential of QMOFs for adsorption-based C 4 H 10 separation from air for the first time in the literature.…”

Section: Introductionmentioning

confidence: 99%

Hierarchical Computational Screening of Quantum Metal–Organic Framework Database to Identify Metal–Organic Frameworks for Volatile Organic-Compound Capture from Air

Ercakir,

Aksu,

Altintas

et al. 2023

ACS Eng. Au

View full text Add to dashboard Cite

The design and discovery of novel porous materials that can efficiently capture volatile organic compounds (VOCs) from air are critical to address one of the most important challenges of our world, air pollution. In this work, we studied a recently introduced metal−organic framework (MOF) database, namely, quantum MOF (QMOF) database, to unlock the potential of both experimentally synthesized and hypothetically generated structures for adsorption-based n-butane (C 4 H 10 ) capture from air. Configurational Bias Monte Carlo (CBMC) simulations were used to study the adsorption of a quaternary gas mixture of N 2 , O 2 , Ar, and C 4 H 10 in QMOFs for two different processes, pressure swing adsorption (PSA) and vacuum-swing adsorption (VSA). Several adsorbent performance evaluation metrics, such as C 4 H 10 selectivity, working capacity, the adsorbent performance score, and percent regenerability, were used to identify the best adsorbent candidates, which were then further studied by molecular simulations for C 4 H 10 capture from a more realistic seven-component air mixture consisting of N 2 , O 2 , Ar, C 4 H 10 , C 3 H 8 , C 3 H 6 , and C 2 H 6 . Results showed that the top five QMOFs have C 4 H 10 selectivities between 6.3 × 10 3 and 9 × 10 3 (3.8 × 10 3 and 5 × 10 3 ) at 1 bar (10 bar). Detailed analysis of the structure−performance relations showed that low/mediocre porosity (0.4−0.6) and narrow pore sizes (6−9 Å) of QMOFs lead to high C 4 H 10 selectivities. Radial distribution function analyses of the top materials revealed that C 4 H 10 molecules tend to confine close to the organic parts of MOFs. Our results provided the first information in the literature about the VOC capture potential of a large variety and number of MOFs, which will be useful to direct the experimental efforts to the most promising adsorbent materials for C 4 H 10 capture from air.

show abstract

ML‐driven models for predicting CO₂ uptake in metal–organic frameworks (MOFs)

Achour,

Hosni

2024

Can J Chem Eng

View full text Add to dashboard Cite

This study advances the discourse on the application of machine learning (ML) algorithms for the predictive analysis of CO2 uptake in metal–organic frameworks (MOFs), with a nuanced focus on the CATBoost model's capability to navigate the complexities inherent in MOFs' heterogeneous landscape. Building upon and extending the comparative analysis, our investigation underscores the CATBoost model's remarkable prediction robustness, characterized by a significant reduction in root mean square error (RMSE) and an enhanced R‐squared (R2) value, thereby affirming its superior accuracy and reliability in forecasting CO2 adsorption. A pivotal aspect of our research is the integration of SHapley Additive exPlanations (SHAP) values for a detailed assessment of feature importance, which not only corroborated ‘pressure’ and ‘surface area’ as pivotal determinants of CO2 uptake but also illuminated the model's advanced analytical capabilities in handling categorical features and mitigating overfitting, even within a dataset marked by intricate and non‐linear patterns. Our quantitative and conceptual analysis, showcasing up to a 15% improvement in RMSE over previous models, reveals the CATBoost model's unparalleled efficiency in discerning the multifaceted interplay of factors influencing CO2 adsorption. This is crucial for the strategic engineering of MOFs with optimized properties. Beyond ‘pressure’ and ‘surface area’, our SHAP analysis highlighted other descriptors with substantial values, elucidating their contributions to CO2 uptake and providing invaluable insights for the MOF design process.

show abstract

Predicting band gaps of MOFs on small data by deep transfer learning with data augmentation strategies

Abstract: Pretrained deep learning models are fine-tuned by our porphyrin-based MOF database using data augmentation strategies to demonstrate how deep transfer learning can predict the properties of MOFs with limited training data.

Cited by 10 publications

References 56 publications

Revolutionizing the structural design and determination of covalent–organic frameworks: principles, methods, and techniques

Revolutionizing the structural design and determination of covalent–organic frameworks: principles, methods, and techniques

Hierarchical Computational Screening of Quantum Metal–Organic Framework Database to Identify Metal–Organic Frameworks for Volatile Organic-Compound Capture from Air

ML‐driven models for predicting CO₂ uptake in metal–organic frameworks (MOFs)

Contact Info

Product

Resources

About

Predicting band gaps of MOFs on small data by deep transfer learning with data augmentation strategies

Abstract: Pretrained deep learning models are fine-tuned by our porphyrin-based MOF database using data augmentation strategies to demonstrate how deep transfer learning can predict the properties of MOFs with limited training data.

Cited by 10 publications

References 56 publications

Revolutionizing the structural design and determination of covalent–organic frameworks: principles, methods, and techniques

Revolutionizing the structural design and determination of covalent–organic frameworks: principles, methods, and techniques

Hierarchical Computational Screening of Quantum Metal–Organic Framework Database to Identify Metal–Organic Frameworks for Volatile Organic-Compound Capture from Air

ML‐driven models for predicting CO2 uptake in metal–organic frameworks (MOFs)

Contact Info

Product

Resources

About

ML‐driven models for predicting CO₂ uptake in metal–organic frameworks (MOFs)