The pursuit of a clean and healthy environment has stimulated much effort in the development of technologies for the utilization of hydrogen-based energy. A critical issue is the need for practical systems for hydrogen storage, a problem that remains unresolved after several decades of exploration. In this context, the possibility of storing hydrogen in advanced carbon materials has generated considerable interest. But confirmation and a mechanistic understanding of the hydrogen-storage capabilities of these materials still require much work. Our previously published work on hydrogen uptake by alkali-doped carbon nanotubes cannot be reproduced by others. It was realized by us and also demonstrated by Pinkerton et al. that most of the weight gain was due to moisture, which the alkali oxide picked up from the atmosphere. Here we describe a different material system, lithium nitride, which shows potential as a hydrogen storage medium. Lithium nitride is usually employed as an electrode, or as a starting material for the synthesis of binary or ternary nitrides. Using a variety of techniques, we demonstrate that this compound can also reversibly take up large amounts of hydrogen. Although the temperature required to release the hydrogen at usable pressures is too high for practical application of the present material, we suggest that more investigations are needed, as the metal-N-H system could prove to be a promising route to reversible hydrogen storage.

Hunting for chemicals with favorable pharmacological, toxicological, and pharmacokinetic properties remains a formidable challenge for drug discovery. Deep learning provides us with powerful tools to build predictive models that are appropriate for the rising amounts of data, but the gap between what these neural networks learn and what human beings can comprehend is growing. Moreover, this gap may induce distrust and restrict deep learning applications in practice. Here, we introduce a new graph neural network architecture called Attentive FP for molecular representation that uses a graph attention mechanism to learn from relevant drug discovery data sets. We demonstrate that Attentive FP achieves state-of-the-art predictive performances on a variety of data sets and that what it learns is interpretable. The feature visualization for Attentive FP suggests that it automatically learns nonlocal intramolecular interactions from specified tasks, which can help us gain chemical insights directly from data beyond human perception.