(412d) Predictions of Adsorption of NH3-Containing Mixtures for Screening of MOFs for Plasma-Assisted NH3 Synthesis and Complementary Energy and Environmental Applications

Ammonia is a critical molecule in the context of energy and sustainability. For instance, NH₃ synthesis is necessary for food security but demands 2% of the world’s energy and emits 2% of global CO₂ emissions. In another instance, NH₃ could play a role in “decarbonizing” the economy through its potentially use as a “hydrogen storage” vehicle within the context of a “hydrogen economy” or as reliable energy carrier on its own. But since NH₃ is already toxic at very low concentrations, there may be also concerns with NH₃ leakage during NH₃ storage, transportation and utilization. Intriguingly, many of the challenges associated with NH₃ production, storage, transport and safety could be addressed by the exploitation of adsorbent materials.

The challenge to “decarbonized” NH₃ production is achieving NH₃ at low pressure (for compatibility for intermittent decentralized green hydrogen feedstock). Pre-activating the N₂ molecules in the feedstock with a plasma can enable synthesis at low temperature, hence removing the need to increase the pressure to circumvent equilibrium limitations. But as the plasma has been shown to also destroy part of the NH₃ product, plasma-assisted synthesis using a porous support that retains and protects the NH₃ product from decomposition has been suggested to improve NH₃ energy yields. On the other hand, one could synthesize NH₃ at the usual high temperature, but remove the need to increase the pressure to circumvent equilibrium limitations by preferentially removing NH₃ through a porous membrane in a membrane reactor or coupling and adsorbent bed to a traditional reactor. In challenges associated with storage and transportation, adsorption could enable storing and transporting NH₃ without lowering the temperature to liquefy it or excessively increasing the pressure to densify it. Furthermore, the NH₃ leakage could be tackled by “traps” that would effectively adsorb leaked NH₃ under dilute conditions in humid environments.

Given the versatility and tunability of metal-organic frameworks (MOFs), which feature a “design space” that could easily reach trillions of potential of MOF designs, these materials have the potential to provide the “just right” structure for each of these applications. But given the overwhelmingly large MOF design space, computational screening is necessary to either i) identify a few “high-promise” designs for which to attempt synthesis and testing or ii) identify clear design rules for optimal materials from data-driven structure-property relationships.

Here, we established a “hierarchical screening” workflow that allowed us to identify promising MOFs and design rules for the above mentioned applications from a “hybrid” database consisting of ~12,000 previously synthesized MOFs (extracted from the CoRE MOF database) and ~3,000 hypothesized MOFs (created using the in-house ToBaCCo-3.0 code). The set up consists of i) a preliminary screening stage using NH₃, N₂, H₂O and Henry constant’s to enable the identification of an initial diverse subset of 300 MOFs for each application-operating condition, ii) a second stage where grand canonical Monte Carlo (GCMC) simulations are used to predict mixture adsorption on the diverse set of MOFs, iii) a third stage where GCMC data for the diverse subset is used to train machine learning (ML) models to predict mixture adsorption in the complete hybrid database, iv) a fourth stage where the top-300 MOFs identified with ML-models is verified with GCMC simulations, and added to the training data. Looping between the third and four stage was done until “convergence” of the top-300 MOF was achieved.

Conditions for the above applications, span temperatures from 300K to 700K, total pressures from 0.1 bar to 30 bar, and NH₃ concentrations in mixtures (NH₃/N₂/H₂ and NH₃/N₂/O₂/Ar) ranging from 25 ppm to 15%. In cases where adsorption under humid conditions was necessary, screening was limited to hydrophobic MOFs (as identified by their low H₂O Henry’s constant). In presenting our results, we i) shed light on the potential use of specific MOFs in addressing practical challenges related to ammonia production, storage, transportation and safety, ii) discuss the “overlap” of promising MOFs across the different applications, iii) provide design rules to maximize the performance of adsorbent materials in general for the above applications through the discussion of structure-property relationships, and iv) discuss the efficacy for data generation and usage to train ML models that solely focus on being good at predicting the properties of the “top” materials for a given application.