Periodic Reporting for period 1 - KNOWSKITE-X (Knowledge-driven fine-tuning of perovskite-based electrode materials for reversible Chemicals-to-Power devices)
Reporting period: 2023-01-01 to 2024-06-30
The main objective of the KNOWSKITE-X project is to discover new oxides embedding various elements to be used as electrodes of solid oxide fuel cells (SOFC) and solid oxide electrolysers (SOEC).
Especially, we target reversible devices capable of operating in both modes, SOFC and SOEC. In fuel cell mode, the device is expected to convert chemical energy into electrical energy. When operating as electrolyser it uses electrical energy to produce chemical fuels. As a result, the reversible operation of these devices enables the integration of intermittent renewable energy sources, such as wind and solar, with the electrical grid by storing the excess energy as chemical fuel.
Today's state-of-the-art electrode materials consist of mixed oxides of yttrium, nickel and zirconium (anode side) and of lanthanum, strontium, cobalt and iron (LSCF) at the cathode side. The key concept of KNOWSKITE-X is to discover new formulations of active mixed oxides with perovskite structure among the billion of possibilities offered by the versatility of such materials class. The targeted materials would be moreover free (or with significantly reduced amount) of toxic and critical content.
To meet this challenge, we follow a science-based methodology boosted by artificial intelligence. In practice, the project pushes towards validating innovative, robust, and use-relevant methodologies based on a smart combination of:
• multiply relevant spectroscopic characterization, including in situ and operando approaches
• multi-scale modelling
• data-enabled knowledge discovery.
Such methodology involves a combination of state-of art preparation of functionalised materials, routine and advanced characterization, machine learning and multi-scale modelling to achieve the discovery of original and relevant scientific knowledge required to sustain the rational design of optimized candidate electrode materials. The targeted knowledge and methods are intended in a wider perspective and a generalisation of the experimental approach will be demonstrated along another user’s case: the development of carbon-based materials for supercapacitors.
Especially, we target reversible devices capable of operating in both modes, SOFC and SOEC. In fuel cell mode, the device is expected to convert chemical energy into electrical energy. When operating as electrolyser it uses electrical energy to produce chemical fuels. As a result, the reversible operation of these devices enables the integration of intermittent renewable energy sources, such as wind and solar, with the electrical grid by storing the excess energy as chemical fuel.
Today's state-of-the-art electrode materials consist of mixed oxides of yttrium, nickel and zirconium (anode side) and of lanthanum, strontium, cobalt and iron (LSCF) at the cathode side. The key concept of KNOWSKITE-X is to discover new formulations of active mixed oxides with perovskite structure among the billion of possibilities offered by the versatility of such materials class. The targeted materials would be moreover free (or with significantly reduced amount) of toxic and critical content.
To meet this challenge, we follow a science-based methodology boosted by artificial intelligence. In practice, the project pushes towards validating innovative, robust, and use-relevant methodologies based on a smart combination of:
• multiply relevant spectroscopic characterization, including in situ and operando approaches
• multi-scale modelling
• data-enabled knowledge discovery.
Such methodology involves a combination of state-of art preparation of functionalised materials, routine and advanced characterization, machine learning and multi-scale modelling to achieve the discovery of original and relevant scientific knowledge required to sustain the rational design of optimized candidate electrode materials. The targeted knowledge and methods are intended in a wider perspective and a generalisation of the experimental approach will be demonstrated along another user’s case: the development of carbon-based materials for supercapacitors.
Modelling activities so far have covered the 2 lower scales of theoretical modelling. More precisely, we have built an original and relevant atomic-scale model of the state-of-art SOFC cathode surface by he means of DFT modelling. In parallel kinetic models have been made, which allow the prediction and modelling of key descriptors such as polarisation curves, prefigurating electrochemical impedance spectroscopy results and operating the way to an in depth understanding of the complex processes at play. First predictions could hence be made on the effect of experimental parameters on the material properties in relevant context. In parallel, we have set up of surrogate models and methods for training the machine learning (ML) algorithms using the actual abundance of unlabelled data on perovskite class materials. Moreover, we are developing strategies for faster modelling using ML and in return, we are developing strategies for handling of missing data needed for ML training by means of DFT modelling. Hence, our results prefigure a genuine DFT-ML hybrid modelling.
Moreover, 3 categories of active materials are being designed, synthesised and experimentally investigated by the consortium:
(i) benchmark materials - the current materials, with slight modifications, as LSCF based cathode materials or carbon-based supercapacitor electrodes. We are investigating current as well as cross-topic characterisation strategies, while evaluating both their repeatability and relevance.
(ii) model materials - as the real life materials are sometime too complex (or too thick, or not transparent enough) to be investigated in depth; In this view, we have designed discrete models: simpler versions, thin films, monocrystals... in order to validate our scientific hypotheses and the results obtained from theoretical modelling.
(iii) candidate materials with emphasis on no or less critical content based on naturally abundant elements such as iron, calcium or titanium.
In particular, new preparations strategies, of which up-scalable paths, are being explored to better control the textural properties of mixed oxides and carbon.
We have successfully characterised most of the as-prepared samples using common routine techniques such as N2-sorption or powder X-Rays diffraction, as well as advanced characterisation such as in situ X-Rays absorption.
At present, a first series of more than one hundred of potential formulations are being experimentally investigated at KNOWSKITE-X partners.
Moreover, 3 categories of active materials are being designed, synthesised and experimentally investigated by the consortium:
(i) benchmark materials - the current materials, with slight modifications, as LSCF based cathode materials or carbon-based supercapacitor electrodes. We are investigating current as well as cross-topic characterisation strategies, while evaluating both their repeatability and relevance.
(ii) model materials - as the real life materials are sometime too complex (or too thick, or not transparent enough) to be investigated in depth; In this view, we have designed discrete models: simpler versions, thin films, monocrystals... in order to validate our scientific hypotheses and the results obtained from theoretical modelling.
(iii) candidate materials with emphasis on no or less critical content based on naturally abundant elements such as iron, calcium or titanium.
In particular, new preparations strategies, of which up-scalable paths, are being explored to better control the textural properties of mixed oxides and carbon.
We have successfully characterised most of the as-prepared samples using common routine techniques such as N2-sorption or powder X-Rays diffraction, as well as advanced characterisation such as in situ X-Rays absorption.
At present, a first series of more than one hundred of potential formulations are being experimentally investigated at KNOWSKITE-X partners.
Significant methodological advances have been made in the ML-enabled materials properties prediction, which have already resulted in several scientific publications.
The DFT model for LSCF surface we have set up is clearly improved as compared to the published ones, and embeds real-life non-symmetrical substitutions and disorder induced by mixing the elements in the final LSCF oxide. As a result, the activation energies we are computing will be more accurate than the existing values.
Improved benchmark-based and new candidate materials have been discovered, and new preparation strategies are being explored and assessed, paving the way to genuinely improved materials for energy applications in general. Our results are under the microscope and will soon be published.
Procedures are being discussed to draw future relevant standards for materials characterisation. At present, some workflows for modelling (MODA) and characterisation (CHADA) processes are being tuned by the involved partners.
The DFT model for LSCF surface we have set up is clearly improved as compared to the published ones, and embeds real-life non-symmetrical substitutions and disorder induced by mixing the elements in the final LSCF oxide. As a result, the activation energies we are computing will be more accurate than the existing values.
Improved benchmark-based and new candidate materials have been discovered, and new preparation strategies are being explored and assessed, paving the way to genuinely improved materials for energy applications in general. Our results are under the microscope and will soon be published.
Procedures are being discussed to draw future relevant standards for materials characterisation. At present, some workflows for modelling (MODA) and characterisation (CHADA) processes are being tuned by the involved partners.