Engineering topics
Publications and source records attributed to Kjær, Emil T. S..
We present MLstructureMining, a machine learning tool that identifies a structural model from an experimental pair distribution function. We show how the method can be used for structure analysis of both crystalline and nanocrystalline materials.
Iridium nanoparticles are important catalysts for several chemical and energy conversion reactions. Studies of iridium nanoparticles have also been a key for the development of kinetic models of nanomaterial formation. However, compared to other metals such as gold or platinum, knowledge on the nature of prenucleation species and structural insights into the resultant nanoparticles are missing, especially for nanoparticles obtained from Ir$_x$Cl$_y$ precursors investigated here. We use in situ X-ray total scattering (TS) experiments with pair distribution function (PDF) analysis to study a simple, surfactant-free synthesis of colloidal iridium nanoparticles. The reaction is performed in methanol at 50 °C with only a base and an iridium salt as precursor. From different precursor salts–IrCl 3 , IrCl 4 , H 2 IrCl 6 , or Na 2 IrCl 6 –colloidal nanoparticles as small as Ir ~55 are obtained as the final product. The nanoparticles do not show the bulk iridium face-centered cubic ($fcc$) structure but show decahedral and icosahedral structures. The formation route is highly dependent on the precursor salt used. Using IrCl 3 or IrCl 4 , metallic iridium nanoparticles form rapidly from Ir$_x$Cl$_y$$^{n–}$ complexes, whereas using H 2 IrCl 6 or Na 2 IrCl 6 , the iridium nanoparticle formation follows a sudden growth after an induction period and the brief appearance of a crystalline phase. With H 2 IrCl 6 , the formation of different Ir$_n$ ($n$ = 55, 55, 85, and 116) nanoparticles depends on the nature of the cation in the base (LiOH, NaOH, KOH, or CsOH, respectively) and larger particles are obtained with larger cations. As the particles grow, the nanoparticle structure changes from partly icosahedral to decahedral. In conclusion, the results show that the synthesis of iridium nanoparticles from Ir$_x$Cl$_y$ is a valuable iridium nanoparticle model system, which can provide new compositional and structural insights into iridium nanoparticle formation and growth.
Characterization of material structure with X-ray or neutron scattering using e.g. Pair Distribution Function (PDF) analysis most often rely on refining a structure model against an experimental dataset. However, identifying a suitable model is often a bottleneck. Recently, automated approaches have made it possible to test thousands of models for each dataset, but these methods are computationally expensive and analysing the output, i.e. extracting structural information from the resulting fits in a meaningful way, is challenging. Our Machine Learning based Motif Extractor (ML-MotEx) trains an ML algorithm on thousands of fits, and uses SHAP (SHapley Additive exPlanation) values to identify which model features are important for the fit quality. We use the method for 4 different chemical systems, including disordered nanomaterials and clusters. ML-MotEx opens for a type of modelling where each feature in a model is assigned an importance value for the fit quality based on explainable ML.
Explore the source record for details and available documents.
A surfactant-free synthesis of precious metal nanoparticles (NPs) performed in alkaline low-boiling-point solvents has been recently reported. Monoalcohols are here investigated as solvents and reducing agents to obtain colloidal Os nanoparticles by using low-temperature (<100 °C) surfactant-free syntheses. The effect of the precursor (OsCl 3 or H 2 OsCl 6 ), precursor concentration (up to 100 mM), solvent (methanol or ethanol), presence or absence of a base (NaOH), and addition of water (0 to 100 vol %) on the resulting nanomaterials is discussed. It is found that no base is required to obtain Os nanoparticles as opposed to the case of Pt or Ir NPs. The robustness of the synthesis for a precursor concentration up to 100 mM allows for the performance of X-ray total scattering with pair distribution function (PDF) analysis, which shows that 1–2 nm hexagonal close packed (hcp) NPs are formed from chain-like [OsO x Cl y ] complexes.