Samenvatting
Introduction -
Zeolites and mesoporous materials are widely used as catalysts or separating agents in the refining and petrochemical industries. The understanding of adsorption phenomena is critical in the comprehension and the modeling of reaction kinetics. Selectivity and adsorption efficiency are often characterized by Henry constants and adsorption enthalpies. It would therefore be advantageous if one could, for a preliminary evaluation, estimate the adsorption properties of mesoporous materials for various hydrocarbons. In this work, partial least square regression has been performed using structural and topological descriptor sets [1] to obtain predictive models for the adsorption properties of mesoporous materials and adsorbate molecules in gas phase.
Experimental -
Adsorption enthalpies and Henry constants used for modeling were experimentally obtained for MCM-41, MCM-48, PHTS, SBA-15 for a set of linear and branched alkanes (20-23 molecules). The training set (10 molecules) was selected by a clustering analysis to contain as much variance as possible from the descriptor set. A structural (number of carbon atoms, Kappa and Chi indices) or topological (number of carbon atoms and branching factors) descriptor set was used. The models were validated by "Leave-One-Out" cross-validation and external validation with the prediction set. Logarithmic values were used for the Henry adsorption constants during the modeling process.
Results and discussion -
Generally, models based on both descriptor sets yielded similar results for adsorption enthalpies but use of topological descriptor sets offers a more straightforward interpretation of the model's coefficients. A clear influence of branching factors in the model's coefficients was observed. Predicting accuracies approaching the experimental error or better were obtained for these non-shape selective materials. The modeling of Henry constants reveals a significantly higher accuracy for the structural descriptor set. Addition of a physicochemical descriptor set (e.g. vapor pressure) enhances the predictability of the models.
Conclusion -
Models established by partial least square regression and few descriptors describe the gas phase adsorption of linear and branched alkanes on mesoporous materials both qualitatively and quantitatively. The use of a topological set of descriptors gives new insights in the influence of branching factor towards selectivity. The presented models are useful with respect to the prediction of adsorption properties for separation or catalytic processes.
Acknowledgements -
Joeri Denayer is grateful to FWO Vlaanderen, Tim Duerinck wishes to thank IFP for its financial and scientific support.
[1] P.Leflaive et al, Microporous and Mesoporous Materials 132 (2010) 246.
E-mail corresponding author: [email protected]
Zeolites and mesoporous materials are widely used as catalysts or separating agents in the refining and petrochemical industries. The understanding of adsorption phenomena is critical in the comprehension and the modeling of reaction kinetics. Selectivity and adsorption efficiency are often characterized by Henry constants and adsorption enthalpies. It would therefore be advantageous if one could, for a preliminary evaluation, estimate the adsorption properties of mesoporous materials for various hydrocarbons. In this work, partial least square regression has been performed using structural and topological descriptor sets [1] to obtain predictive models for the adsorption properties of mesoporous materials and adsorbate molecules in gas phase.
Experimental -
Adsorption enthalpies and Henry constants used for modeling were experimentally obtained for MCM-41, MCM-48, PHTS, SBA-15 for a set of linear and branched alkanes (20-23 molecules). The training set (10 molecules) was selected by a clustering analysis to contain as much variance as possible from the descriptor set. A structural (number of carbon atoms, Kappa and Chi indices) or topological (number of carbon atoms and branching factors) descriptor set was used. The models were validated by "Leave-One-Out" cross-validation and external validation with the prediction set. Logarithmic values were used for the Henry adsorption constants during the modeling process.
Results and discussion -
Generally, models based on both descriptor sets yielded similar results for adsorption enthalpies but use of topological descriptor sets offers a more straightforward interpretation of the model's coefficients. A clear influence of branching factors in the model's coefficients was observed. Predicting accuracies approaching the experimental error or better were obtained for these non-shape selective materials. The modeling of Henry constants reveals a significantly higher accuracy for the structural descriptor set. Addition of a physicochemical descriptor set (e.g. vapor pressure) enhances the predictability of the models.
Conclusion -
Models established by partial least square regression and few descriptors describe the gas phase adsorption of linear and branched alkanes on mesoporous materials both qualitatively and quantitatively. The use of a topological set of descriptors gives new insights in the influence of branching factor towards selectivity. The presented models are useful with respect to the prediction of adsorption properties for separation or catalytic processes.
Acknowledgements -
Joeri Denayer is grateful to FWO Vlaanderen, Tim Duerinck wishes to thank IFP for its financial and scientific support.
[1] P.Leflaive et al, Microporous and Mesoporous Materials 132 (2010) 246.
E-mail corresponding author: [email protected]
| Originele taal-2 | English |
|---|---|
| Titel | 1st International Workshop of the European Nanoporous Materials Institute of Excellence, 4-5 October, 2010, Antwerpen, Belgium |
| Status | Published - 5 okt 2010 |
| Evenement | Unknown - Stockholm, Sweden Duur: 21 sep 2009 → 25 sep 2009 |
Conference
| Conference | Unknown |
|---|---|
| Land/Regio | Sweden |
| Stad | Stockholm |
| Periode | 21/09/09 → 25/09/09 |