Accelerated DFT-Based Design of Materials for Ammonia Storage

Abstract

Future energy carriers are needed in order to lower the CO2 emissions resulting from the burning of fossil fuels. One possible energy carrier is ammonia, which can be stored safely and reversibly in metal halide ammines; however, the release often occurs in multiple steps at too high temperatures. Therefore, there is a need for new materials, releasing the ammonia in a narrow temperature interval. To search for new mixed metal halide chlorides, we use DFT calculations guided by a genetic algorithm (GA) to expedite the search, as the defined search space allowing up to three different metals contains more than 100,000 different structures. Here, we search for materials releasing the ammonia between 0 and 100 °C, a temperature range suitable for system integration with low-temperature polymer electrolyte membrane fuel cells (PEMFC). The efficiency of the implemented algorithm is verified by three trial runs capable of finding the same optimal mixtures starting from different random populations, testing < 5% of the candidates. Some of the best candidates are already confirmed experimentally and others offer a record high, accessible hydrogen capacity exceeding 9 wt%. Among the identified materials is the first known high-capacity ternary metal halide ammine, which we have subsequently synthesized and confirmed the ammonia storage properties using temperature programmed desorption (TPD). IntroductionAmmonia is a widely used chemical with many different applications; most importantly as the main component in fertilizers, sustaining the growth of the world's population. 1 A major drawback of ammonia is, however, its toxicity, requiring careful handling and storage. 2 It is well-known that pure metal halide ammines have the ability to store ammonia, and release it when heat is supplied. 3,4 These solid storage materials have been proposed as energy carriers for transportation usage 5-12 and for on-board reduction of NOX gases from combustion processes. 13 The metal halide ammines often display good kinetics 14,15 , but most of the pure metal halides release the ammonia at high temperatures and in multiple steps spread over broad temperature ranges. 8,16 Therefore, new materials with optimized release patterns are needed if ammonia should be used as a future energy carrier for transportation. A number of mixed metal halide ammines have already been investigated, both mixing the cations and the anions 17-19 , resulting in materials performing better than their individual components, encouraging the search for other mixed materials.

Ligand-Field-Dependent Behavior of Meta-GGA Exchange in Transition-Metal Complex Spin-State Ordering

“…Approximate density functional theory (DFT) remains the method of choice for computational discovery [4][5][6][7][8][9][10][11] , owing to its favorable balance of accuracy and efficiency. 12 Nevertheless, delocalization errors [13][14] and other biases in semi-local exchange approximations [15][16] (e.g., the generalized gradient approximation or GGA) produce systematic biases toward low-spin states [17][18] that prevent prediction of either qualitative (i.e., ground state identity) or quantitative (i.e., energetic splitting between states) spin-state properties.…”

Computational Discovery of Hydrogen Bond Design Rules for Electrochemical Ion Separation

“…In response to recent experimental and theoretical advances, the DFT-based computational screening has become an increasingly valuable tool for the design and discovery of new materials [66][67][68][69][70][71] thanks both to recent improvements in computational efficiency and accuracy (e.g., in descriptions of intermolecular forces through direct treatment of dispersion [72][73]. In this work, we carry out the first computational screen to identify design strategies for Fc + core functionalization to maximize formate selectivity and reversibility in the Fc/Fc + redox system.…”

Leveraging Cheminformatics Strategies for Inorganic Discovery: Application to Redox Potential Design

“…Unsigned average error in bond length prediction (in Å) compared against a DFT B3LYP optimized structure over representative bidentate(5) and representative monodentate(5) octahedral Fe(II) and Fe(III) low-spin and high-spin complexes from the UFF (red bars), metalligand database (grey bars), and an ANN (blue bars).…”