WP1 performs materials screening. The project has focused on setting up the infrastructure, new algorithms and methods, and workflows. This has required us to develop a software framework at the heart of the machine learning process. We have established a flexible database format for collecting data and converting it to a convenient, portable form. Nanolayers has developed machine-learning tools to characterize known catalyst materials with the aim of designing novel ones. The CritCat software libraries include several well-established machine-learning tools (neural networks, genetic algorithm) and cutting edge methods, all optimized to describe the materials of interest.
We rationalize the information from other WPs by developing an initial set of physical parameters (descriptors) and benchmark them against established reaction information. We have optimized workflows for the DFT simulations to collect reference datasets for machine learning. The data set of small TM clusters has been used for the training a neural-network (NN) force field, and model has been expanded to include total energies. We can now generate model structures for bimetallic clusters very efficiently.
Aalto has made considerable advances in finding suitable reaction descriptors for the HER reaction. Based on DFT calculations and new algorithms, we are able to efficiently predict HER activity of selected catalyst systems.
In WP2 for synthesis, UOB/Swansea has produced size-selected elemental Au, Pd, Pt, Ag, and Co clusters as well as bimetallic clusters. Swansea has also produced size-selected MoS2 clusters (with Ni/Co doping or additional S). INL has prepared an extensive set of TM-based phosphide and carbide electrodes. NPL has investigated the synthesis of TM chalcogenide and phosphide nanoparticles under electrochemical conditions. Tethis has focused on its key technologies around flame spray pyrolysis, and samples of PGM nanopowders nanostructured layers have been provided for the other partners in WP3-4.
Scale-up of cluster beam production of nanoparticles within WP2 has been led by Swansea. Quantitites of cluster catalyst on the mg to gram scale are now available, based on the Matrix Assembly Cluster Source (MACS).
WP3 focuses on characterization. By applying various spectroscopic methods and electrochemical test, we have characterized MoS2 and Ni-MoS2 clusters, PGM nanoparticles, Pt-M nanoparticles (M = Ti, Cu, Ni, etc.), Ni2P nanoparticles, Ni2P and Ni5P4 electrodes, MoS2 films, etc. Altogether, the results in WP3 demonstrate our capability to perform full-scale characterization of the catalyst materials synthesized. The experimental cluster sizes correspond to few hundred atoms to few thousand atoms, and they can be simulated by DFT, as demonstrated by TUT.
WP4 considers catalyst performance. NPL, INL, and Syngaschem have performed tests for the PGM and TM nanoparticle catalysts produced in WP2. TUT and Aalto have focused on selected gas phase reactions, and electrochemical HER and OER reactions. While we have developed our experimental expertise further and made scientific findings, the obtained results have also provided us important reference data for materials screening (WP1).
TUT has investigated selected gas phase reactions on PGM-based catalysts by DFT. Aalto has modeled the electrochemical HER reaction on MoS2 catalysts. Similar simulations are being carried out on TM phosphide systems. The simulation data has been uploaded on the CritCat server.
WP5 aims for implementing promising materials (defined by WP4) into electrochemical applications at the device level, with full or partial PGM replacement. Significant knowledge has been obtained which allows improvements in the performance of electrochemical devices for the production of hydrogen and/or synthetic fuels. Pilot-scale testing was realized for (i) full PGM replacement at the cathode of real PEM water electrolyzers and CO2 electrochemical converters and (ii) efficient PGM utilization in PEM alcohol-assisted water electrolyzers.
