Leveraging composition-based energy material descriptors for machine learning models

Trezza, Giovanni; Chiavazzo, Eliodoro

doi:10.1016/j.mtcomm.2023.106579

Cited by 4 publications

(7 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These fruitful regions of the parameter space can be identified and visualized by reducing the five variables into a lower number of features following the recently introduced Control Group Feature analysis (herein, the two mixed features are defined as x 1 and x 2 in Figure S17; details in Experimental Section). 20 The approach is inspired by Buckingham analysis and based on the principle that variable combinations that offer the same mixed feature value will offer comparable performance by exploiting alternate maxima in the parameter space. This analysis highlights regions of the parameter matrix containing high performance conditions as clustered highperforming values in scatter plots.…”

Section: ■ Results and Discussionmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

“…The feature grouping separates values into classes based on a threshold for the corresponding properties, i.e., above (class 1) and below (class 0) the threshold following the methodology recently introduced by Trezza & Chiavazzo. 20 Such thresholds are −2.2 for y obj 1 , 1.3 μmol for CO, 300 for TON CO , 18.5 min −1 for TOF CO , 0.14 for QY CO . In particular, given the original set of 5 features (x 1 , ..., x 5 ), the corresponding dimensionless quantities are denoted (x1, ..., x5), where x 1 { } × being a matrix computed by means of a multiobjective optimization criterion in two steps.…”

Section: ■ Conclusionmentioning

confidence: 98%

“…Acquiring a systematic data set that maps the parameter space then enabled the application of machine learning algorithms to extract relationships between the dependent and the independent variables. Furthermore, the role and importance of different parameters on the overall performance was revealed using Shapley additive explanations (SHAP). − Finally, the problem dimensionality was reduced and parameter regions likely to further enhance system performance are exposed by controlling feature grouping . Thereby, the system was holistically optimized and deeper insight into the ruling parameters for each figure of merit was extracted.…”

Section: Introductionmentioning

confidence: 99%

“…17−19 Finally, the problem dimensionality was reduced and parameter regions likely to further enhance system performance are exposed by controlling feature grouping. 20 Thereby, the system was holistically optimized and deeper insight into the ruling parameters for each figure of merit was extracted.…”

Section: ■ Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Multi-Variable Multi-Metric Optimization of Self-Assembled Photocatalytic CO₂ Reduction Performance Using Machine Learning Algorithms

Bonke,

Trezza,

Bergamasco

et al. 2024

J. Am. Chem. Soc.

Self Cite

View full text Add to dashboard Cite

The sunlight-driven reduction of CO 2 into fuels and platform chemicals is a promising approach to enable a circular economy. However, established optimization approaches are poorly suited to multivariable multimetric photocatalytic systems because they aim to optimize one performance metric while sacrificing the others and thereby limit overall system performance. Herein, we address this multimetric challenge by defining a metric for holistic system performance that takes multiple figures of merit into account, and employ a machine learning algorithm to efficiently guide our experiments through the large parameter matrix to make holistic optimization accessible for human experimentalists. As a test platform, we employ a five-component system that self-assembles into photocatalytic micelles for CO 2 -to-CO reduction, which we experimentally optimized to simultaneously improve yield, quantum yield, turnover number, and frequency while maintaining high selectivity. Leveraging the data set with machine learning algorithms allows quantification of each parameter's effect on overall system performance. The buffer concentration is unexpectedly revealed as the dominating parameter for optimal photocatalytic activity, and is nearly four times more important than the catalyst concentration. The expanded use and standardization of this methodology to define and optimize holistic performance will accelerate progress in different areas of catalysis by providing unprecedented insights into performance bottlenecks, enhancing comparability, and taking results beyond comparison of subjective figures of merit.

show abstract

Section: ■ Results and Discussionmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: ■ Conclusionmentioning

confidence: 98%

Section: Introductionmentioning

confidence: 99%

Section: ■ Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Multi-Variable Multi-Metric Optimization of Self-Assembled Photocatalytic CO₂ Reduction Performance Using Machine Learning Algorithms

Bonke,

Trezza,

Bergamasco

et al. 2024

J. Am. Chem. Soc.

Self Cite

View full text Add to dashboard Cite

show abstract

Optimizing MOF properties for seasonal heat storage: a machine learning approach

Trezza,

Bergamasco,

Fasano

et al. 2024

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

In the quest to enhance thermochemical energy storage using promising sorbents, this work presents a study on the optimization of Metal Organic Frameworks (MOFs) properties for gas sorption, with a focus on CO2 and H2O adsorption. Through the analysis of crystallographic descriptors, the study aims to streamline the selection of MOFs that could potentially exceed the performance of existing water sorbent pairs. A comprehensive comparison of sequential learning (SL) algorithms reveals a method for identifying the minimal set of descriptors that influence adsorption properties of MOFs. The protocol involves constructing and training machine learning (ML) models to determine the number of influential descriptors and utilizing SHAP analysis to evaluate their importance. Findings suggest that including only these critical descriptors in the exploration space reduces computational load. Notably, the COMBO and the FUELS algorithms consistently outshine random guessing, validating their efficacy in materials optimization. The challenge of accessing full adsorption properties across the entire coverage range is addressed by a computational screening procedure requiring minimal input data. This method suggests that some vanadium based MOFs, originally designed for different purposes, could surpass the current leading compounds for thermal energy storage, primarily due to their optimal Henry coefficient values for water adsorption.

show abstract

Multi-Variable Multi-Metric Optimisation of Self-Assembled Photocatalytic CO2 Reduction Performance using Machine Learning Algorithms

Bonke,

Trezza,

Bergamasco

et al. 2023

Preprint

View full text Add to dashboard Cite

The sunlight-driven reduction of CO2 into fuels and platform chemicals is a promising approach to enable a circular economy. However, established optimisation approaches are poorly suited to multi-variable multi-metric photocatalytic systems because they aim to optimise one performance metric while sacrificing the others and thereby limit overall system performance. Herein, we address this multi-metric challenge by defining a metric for holistic system performance that takes all figures of merit into account, and employ a machine learning algorithm to efficiently guide our experiments through the large parameter matrix to make holistic optimisation accessible for human experimentalists. As a test platform, we employ a five-component system that self-assembles into photocatalytic micelles for CO2-to-CO reduction, which we experimentally optimised to simultaneously improve yield, quantum yield, turnover number, and frequency while maintaining high selectivity. Leveraging the dataset with machine learning algorithms allows quantification of each parameter’s effect on overall system performance. The buffer concentration is unexpectedly revealed as the dominating parameter for optimal performance, and is nearly four times more important than the catalyst concentration. The expanded use and standardisation of this methodology to define and optimise holistic performance will accelerate progress in different areas of catalysis by providing unprecedented insights into performance bottlenecks, enhancing comparability, and taking results beyond comparison of subjective figures of merit.

show abstract

Leveraging composition-based energy material descriptors for machine learning models

Cited by 4 publications

References 28 publications

Multi-Variable Multi-Metric Optimization of Self-Assembled Photocatalytic CO₂ Reduction Performance Using Machine Learning Algorithms

Multi-Variable Multi-Metric Optimization of Self-Assembled Photocatalytic CO₂ Reduction Performance Using Machine Learning Algorithms

Optimizing MOF properties for seasonal heat storage: a machine learning approach

Multi-Variable Multi-Metric Optimisation of Self-Assembled Photocatalytic CO2 Reduction Performance using Machine Learning Algorithms

Contact Info

Product

Resources

About

Leveraging composition-based energy material descriptors for machine learning models

Cited by 4 publications

References 28 publications

Multi-Variable Multi-Metric Optimization of Self-Assembled Photocatalytic CO2 Reduction Performance Using Machine Learning Algorithms

Multi-Variable Multi-Metric Optimization of Self-Assembled Photocatalytic CO2 Reduction Performance Using Machine Learning Algorithms

Optimizing MOF properties for seasonal heat storage: a machine learning approach

Multi-Variable Multi-Metric Optimisation of Self-Assembled Photocatalytic CO2 Reduction Performance using Machine Learning Algorithms

Contact Info

Product

Resources

About

Multi-Variable Multi-Metric Optimization of Self-Assembled Photocatalytic CO₂ Reduction Performance Using Machine Learning Algorithms

Multi-Variable Multi-Metric Optimization of Self-Assembled Photocatalytic CO₂ Reduction Performance Using Machine Learning Algorithms