Computational Design of New Materials for Ammonia Storage

Many renewable energy technologies are currently being developed, but most have the major drawback, that the energy is not generated when needed, and storage of the energy is therefore essential. For transportation applications,an energy carrier offering a high energy density is needed. Ammonia is one possible energy carrier offering high gravimetric and volumetric capacities.

A major drawback regarding ammonia is however its toxicity requiring careful handling and storage.Instead of handling the ammonia directly, it is possible to store it safely and completely reversibly in many different metal halide salts. The release temperatures vary significantly for the different materials and the release is often a multi-step reaction occurring over a broad temperature interval.

The object of the present thesis has therefore been to design new mixed materials with improved release characteristics using Density Functional Theory (DFT), which offers a predictive accuracy at a limited computational cost. Initially strontium chloride, which is the state-of-the-art material for solid state ammonia storage,was studied in details.

From the gained insights new solid solutions of strontium and barium chlorides were predicted to offer superior storage capacities,which subsequently were experimentally verified.Inspired by the impressive properties offered by the binary mixtures an extensive computational screening was initiated searching for bi- and ternary metal halides.

A large number of metals were allowed, resulting in search spaces containing many thousands of possible candidates, which cannot all be studied using DFT calculations due to consideration of the cumulative computational cost. Therefore a genetic algorithm (GA) was implemented to guide the search more effectively.

To further decrease the need for computational resources a template based screening approach was used to set up the structures.An initial test of the algorithm was performed searching for the mixed hexa ammine offering the highest storage capacity. The investigated search space contained almost 27 000 structures.

By using the GA it was possible to identifythe global optimum structure in three consecutive runs starting from differentrandom populations, examining less than two percent of the candidates.Following the successful initial test of the algorithm a more advanced screeningwas initiated. The complexity was increased by including checks of multiple intermediateammine phases to determine the correct release patterns, resultingin a total of more than 100 000 structures to test.

The objective of the searchwas to find optimised materials releasing most possible ammonia in a narrowtemperature interval. Again the robustness of the algorithm was confirmed byrerunning the algorithm, which showed a very high consistency in identifyingthe best candidates.Interestingly, the algorithm correctly identified all the barium-strontium mixtureswhich were already known to exist from the previous studies as goodcandidates.

Furthermore a ternary alkaline-earth mixture, Ba4CaSr3Cl16 wasidentified, and the predicted release was verified experimentally. The two bestfound candidates, Ca4Cu2Y2Cl16 and Sr4Cu2Y2Cl16, are examples of materials,one would not suggest using chemical intuition. This shows that thealgorithm is able to identify new materials by using the available operators,which are based on a combination of chemical knowledge and random events.The initial experimental tests of the identified candidates showed promisingresults, but did also reveal that the materials can be challenging to synthesize,and might degrade during ammonia desorption due to unwanted side reactions.