Skip to main content

Modelling for the search for new active materials for redox flow batteries

Periodic Reporting for period 1 - SONAR (Modelling for the search for new active materials for redox flow batteries)

Reporting period: 2020-01-01 to 2021-04-30

SONAR will develop a framework for the simulation-based screening of electroactive materials for aqueous and nonaqueous organic redox flow batteries (RFBs). It will adopt a multiscale modelling paradigm, in which simulation methods at different physical scales will be further advanced and linked by combining physics- and data-based modelling. For the iterative traversal of the different scales, exclusion criteria like solubility, standard potentials and kinetics will be defined, and the results for individual candidates will be stored in a database for further processing. To increase the throughput of the screening, SONAR will exploit advanced data integration, analysis and machine-learning techniques, drawing on the growing amount of data produced during the project. The models will be validated e.g. by comparison with measurements of redox potentials for known chemistries, or measurement data of RFB half-cells and lab-sized test cells.
The overall objectives are:
1. to optimize scale models and adapt them to the requirements of organic RFBs
2. to link the models from objective 1 into an automated, high-throughput, multiscale workflow
3. to validate and exploit the developed models
With the start of the project on 01.01.2020 work began in the areas of atomistic and meso-scale, 1D and 3D cell modeling, as well as the linking of the individual subscales. As planned, developments in the areas of stack and system modeling and cost analysis began one year later.
With the aid of quantum mechanical calculations, redox potentials were determined and then compared with tabulated measured values. For various molecules from the quinone range, the calculated values were demonstrated to be highly accurate at different pH values. The results will be presented in an upcoming scientific publication.
The molecular properties of active materials have been defined, which will be simulated using quantum mechanical methods and stored in a database. A first computational protocol has been developed and applied for the calculation of ~4200 reaction energies that will be used to train a machine learning model to predict potential.
An on-lattice kinetic Monte Carlo (kMC) model has been developed which focusses on simulating four types of events: the molecule and ion displacement, the charge/discharge of methyl viologen, the adsorption and desorption of water molecules (Frumkin effect), and the degradation of active material.
Since month 6 of the project, the experimental kinetic characterization of 4-HO-TEMPO has been carried out. Methods for the extraction of the diffusion coefficient (D) and the kinetic constant k° have already been implemented. Two different methods were used for the diffusion coefficient: cyclic voltammetry with Randles-Sevcik formalism, and a rotating electrode with the Levich equation and the Koutecky-Levich approach.
A 0-D U-I-SoC aqueous organic RFB cell model was developed for efficient performance predictions. The model allows the cell voltage and power density to be expressed in terms of the state of charge and electric current density. To establish an up-scaling procedure from the pore-scale structure, electrolyte transport and electrochemical reactions to effective macro-scale parameters for macrohomogeneous descriptions of porous electrodes, periodic microstructures based on simplified geometries were studied.
A micro structure analysis including computational and experimental investigations of different commercially available carbon-based electrode materials was carried out. The structure on the micrometre scale was obtained through the reconstruction of 3D X-ray micro-computed tomography scans with a voxel size in the one-digit micrometer range. A 3D battery simulation tool within the open source platform OpenFOAM was extended to deal with the RFB principle in general and the novel organic substance system in particular.
An electrical circuit model was used to create the basis for electrochemical stack modeling. The resistive network can not only take into account and combine the separate characteristics of the individual cells, but also account for specific stack characteristics such as electrical cross currents. An electrochemical cell was modeled and simulated based on a vanadium RFB to provide a basis for further development and investigation. Using this cell, investigations were carried out on the current density at different flow rates.
An optimized techno-economic model was created in order to be able to take into account further factors influencing the costs and technical properties of organic RFBs. Furthermore, an inventory of existing technical components was made based on laboratory cells as a reference for commercial batteries. To obtain experimental data, different RFBs based on vanadium as a reference and with methylviologen and 4-hydroxy TEMPO were set up and investigated using different electrochemical methods.
A standalone python tool has been implemented, documented and tested, which allows a flexible filtering sequence to be composed with predefined or user-defined functions in order to preselect compounds merely based on their topology. A routine constructing redox-partners from a given compound is under development, in tandem with user-defined SMARTS reaction templates. The redox couples are input to either through quantum mechanics calculations or prediction by a surrogate model. A trained neural network (to be submitted for publication) predicts heats of reaction/formal potentials for the generated redox pairs.
A range of dissemination channels have been used to create awareness about the project’s activities and results: a website ( newsletter, Twitter and LinkedIn. On the initiative of SONAR coordinator Fraunhofer ICT, the Network of (EU-funded) Flow Battery Initiatives “FLORES” was founded in 2021. Its goal is scientific exchange, joint outreach and joint dissemination of project results. Several scientific publications are in preparation for submission, and all project partners are contributing to a book, edited by the SONAR coordinator.
The developments made in SONAR beyond the state of the art were 1) individual scale models and 2) the development of a high-throughput screening. The approach of SONAR is the linear linking of optimized individual scale models and the transfer of results into the respective other scales to calculate the performance values and costs of organic RFBs. So far, the foundations have been laid by creating a data structure and database and starting to train a machine-learning program based on database values. In addition, individual scale models have been developed with the requirements of aqueous organic RFBs. These include, in particular, the consideration of pH changes in quantum mechanical calculations of redox potentials, the development of a kinetic Monte Carlo model for organic RFBs, the development of a fast 0D cell model and the consideration of the specific properties of felts in the modeling of cells. Furthermore, experimental work was carried out to validate and provide experimental data at practical concentrations of active materials.
This work will be continued and intensified until the end of the project. With the start of the work in the area of stack and system modeling as well as techno-economics, an analysis of the costs will be possible for the first time, in addition to the calculation of additional system performance parameters.