SOURCE-LINKED INTELLIGENCE

Machine learning guided discovery of potassium-selective porous silicates

CORDIS · observation · Publication date unknown

Machine learning guided discovery of potassium-selective porous silicates CLEANSE replaces the time-consuming, expensive, and error-prone trial-and-error approach for developing cation-selective sorbents with a predictive, iterative computation-to-experiment loop. First, we will resolve why sodium zirconium cyclosilicate (ZS-9) is highly selective for K+ by training equivariant machine-learning interatomic potentials (MLIPs) on ab initio data and running long (biased) molecular-dynamics simulations under realistic aqueous, multi-ion conditions. These models will also predict experimental observables such as solid-state NMR tensor and Born effective charges for effective NMR and IR spectral prediction. Second, we will perform high-throughput screening of broad synthesizable porous silicate databases, ranking candidates by adsorption capacity, competitive selectivity against Na+/NH4+/H3O+,

Read original source ↗ Open in workspace

recordType

award

status

SIGNED

region

EU

value

207183.12

unit

EUR