Time and place
Monday, 18 May 2026, at 13:00, Bldg. 341, Aud. 022.
Principal supervisor
Associate Professor Andrea Crovetto, DTU Nanolab
Co-supervisor
Professor Eugen Stamate, DTU Nanolab
Assessment committee
Professor Jakob Birkedal Wagner, DTU Nanolab
Professor Julien Bachmann, Friedrich-Alexnder University
Associate Professor Jonathan Staaf Scragg, Uppsala Universitet
Moderator at defence
Senior Researcher Alice Bastos da Silva Fanta, DTU Nanolab
Abstract
In a rapidly changing world where data is collected from every device you own it is surprising to realize that scientific publishing in materials research still works pretty much the same as a hundred years ago and does not care about data itself but only about articles. This problem is amplified by the recent acceleration of experiments and characterization. These so-called high-throughput approaches enable the synthesis of hundreds of materials simultaneously instead of one at a time.
This work has therefore two focus points - materials discovery and how to publish the materials data collected during this work.
The experimental aspect of this work aimed at exploring a new class of materials containing phosphorus, sulfur and one metal here copper. These materials are predicted to be interesting for applications in photovoltaics and catalysis. Using a high-throughput synthesis method, I successfully synthesized two materials (Cu3PS4 and Cu7PS6) in the form of a continuous layer with a few hundred nanometers in thickness. These materials turned out to be very stable semiconductors making them suitable candidates for solar driven catalysis. The detailed results were published in scientific articles.
The remaining data not included in those publications would traditionally not be published, as journals publish only success stories. That means we lose a lot of knowledge about what does not work, leading scientists to repeat failed experiments over and over. This is not only a waste of time but also a missed opportunity as coherent extensive experimental datasets enable machine-learning-based predictions and fundamental physical and chemical insights into materials.
Therefore, the second focus of this thesis was to develop a database to organize all collected experimental data. To make it user friendly and easy to adapt to for other scientists we expanded on an established database for materials simulations. The unique infrastructure enables automation of data analysis as well as publishing of the whole dataset including the failed experiments. This means that all 5,000 data points collected during this project will be available for future scientific use and can contribute to knowledge generation beyond my project.
