The total angular momentum of a closed system is a conserved quantity, which should remain constant in time for any excitation experiment once the pumping signal has been extinguished. Such conservation, however, is never satisfied in practice in any real-time first-principles description of the demagnetization process. Furthermore, there is growing experimental evidence that the same takes place in experiments. The missing angular momentum is usually associated to lattice vibrations, which are not measured experimentally and are never considered in real-time simulations. Here we critically analyze the issue and conclude that current state-of-the-art simulations violate angular momentum conservation already at the electronic level of description. This shortcoming originates from an oversimplified description of the spin-orbit coupling, which includes atomic contributions but neglects completely that of itinerant electrons. We corroborate our findings with time-dependent simulations using model tight-binding Hamiltonians, and show that indeed such conservation can be reintroduced by an appropriate choice of spin-orbit coupling. The consequences of our findings on recent experiments are also discussed.

In recent times, transformer networks have achieved state-of-the-art performance in a wide range of natural language processing tasks. Here we present a workflow based on the fine-tuning of BERT models for different downstream tasks, which results in the automated extraction of structured information from unstructured natural language in scientific literature. Contrary to existing methods for the automated extraction of structured compound-property relations from similar sources, our workflow does not rely on the definition of intricate grammar rules. Hence, it can be adapted to a new task without requiring extensive implementation efforts and knowledge. We test our data-extraction workflow by automatically generating a database for Curie temperatures and one for band gaps. These are then compared with manually curated datasets and with those obtained with a state-of-the-art rule-based method. Furthermore, in order to showcase the practical utility of the automatically extracted data in a material-design workflow, we employ them to construct machine-learning models to predict Curie temperatures and band gaps. In general, we find that, although more noisy, automatically extracted datasets can grow fast in volume and that such volume partially compensates for the inaccuracy in downstream tasks.