PNNL

The Science

Understanding how bulk nitric acid solutions and rare-Earth ions dissolved in these solutions behave at the molecular level is an essential step toward predicting the complex solution chemistry that governs the separation and recovery of critical materials. The team took the large, challenging problem and separated it into smaller pieces in a pair of papers, developing and testing models to compile a complete and consistent physical picture. For nitric acid, researchers developed a custom-trained machine learning interatomic potential (MLIP) that captures proton transfer. They showed the strong acid becomes progressively less dissociated as concentration increases, accompanied by major changes in the local solvation environment. For lanthanides, they focused on the dilute limit system. The team connected traditional quantum-chemistry solvation models with explicit-solution condensed-phase simulations and found that nitrate association is inseparable from hydration-shell reorganization.

The Impact

Predicting how molecular speciation changes as acidity, concentration, and metal–ligand interactions change is a fundamental challenge in critical materials chemistry. Because the full lanthanide–nitric-acid problem is highly complex, researchers are breaking down the problem to develop the necessary understanding in stages and testing whether different molecular models give a consistent physical picture. The team found that universal MLIPs can reproduce a few individual properties but cannot provide a consistently accurate and quantitative description across structure, density, and dissociation yet. Overall, the results of these two studies identify both the molecular physics that must be captured and the limits that must be overcome to predictively model lanthanide speciation in concentrated nitric acid.

Summary

In complex solutions of ions, predicting molecular species remains challenging as a wide range of factors can affect ionic behavior. Researchers used a combination of quantum-mechanical calculations, molecular simulations, and AI to study different pieces of electrolyte speciation. The team examined two complementary pieces of this broader speciation problem: nitric acid solutions across a wide range of concentrations and lanthanide–nitrate chemistry in dilute aqueous solution. They found that overall, molecular speciation depends strongly on the surrounding solution environment. For nitric acid, they developed system-specific MLIP trained on quantum-mechanical data. They could then follow proton transfer and changes in molecular structure over a wide range of concentrations. Simulations then showed that nitric acid becomes a weaker acid at high concentrations. For the lanthanide–nitrate systems, they connected a traditional quantum-chemistry approach that focuses on the metal ion and its nearest water molecules with larger molecular simulations using general-purpose, or universal, MLIPs that explicitly represent the surrounding aqueous solution. They could then examine how nitrate binds to lanthanide ions, how the surrounding water reorganizes, and how stable those associations are in dilute solution. The universal MLIPs captured some aspects of this molecular behavior but did not yet reliably predict the strength of nitrate association. The calculations also revealed a flexible coordination landscape in which nitrate adopts competing monodentate and bidentate binding motifs as water reorganizes around the metal ion. Together, these studies provide molecular benchmarks for two pieces of the larger speciation problem and show why predicting lanthanide chemistry in concentrated nitric acid requires treating acidity, hydration, ligand association, and concentration together. 

Contact

Hadi Dinpajooh, Pacific Northwest National Laboratory, hadi.dinpajooh@pnnl.gov 

Funding

10.1063/5.0303907 – M.D. was supported under Grant No. FWP 85666, a U.S. Department of Energy (DOE), Office of Science (SC), Early Career Research Program award in the Basic Energy Sciences (BES), Chemical Sciences, Geosciences, and Biosciences (CSGB) Division, Condensed Phase and Interfacial Molecular Science (CPIMS) program, to develop machine-learning interatomic potentials for electrolyte solutions. A.M.R., M.D.L., S.E.M., N.J.H., and M.D. were also supported by the National Security Directorate Mission Seed Initiative, under the Laboratory Directed Research and Development (LDRD) program at Pacific Northwest National Laboratory (PNNL), to develop a multiscale approach for determining the thermodynamic parameters for nuclear fuel dissolution. D.M.R. acknowledges support from the DOE BES CSGB Division under the Center for Scalable Predictive Methods for Excitations and Correlated Phenomena, funded as part of the Computational Chemical Sciences program, Grant No. FWP 70942, for the XPS simulation methodology and implementation. C.J.M. acknowledges support by U.S. DOE, Office of Science, BES, CSGB Division, CPIMS program under Grant No. FWP 16249 at PNNL for developing theoretical and computational techniques to simulate molecular phenomena. PNNL is operated by Battelle for the U.S. DOE under Contract No. DE-AC05-76RL01830. Computing resources were generously allocated by PNNL’s Institutional Computing program. This research also used resources of the National Energy Research Scientific Computing Center (NERSC), a DOE SC user facility using NERSC under Award No. BES-ERCAP-26557. The authors thank Niranjan Govind, Jenna A. Pope, and Gregory K. Schenter at PNNL for helpful discussions.

10.1039/d6cp00838k – M.D. was supported under FWP 85666, a DOE SC Early Career Research Program award in the BES CSGB Division, CPIMS program, for developing molecular-scale and reduced models of solutions for critical materials, including DFT and AI-based modeling approaches. A.M.R, M.D., and N.U. were also partially supported by the Generative AI for Science, Energy, and Security Science & Technology Investment under the LDRD program at PNNL, which concluded in fiscal year 2025, to develop multiphysics models of nuclear forensics chemistry. PNNL is a multiprogram national laboratory operated by Battelle for the DOE. This work was also supported by the Center for AI and Center for Cloud Computing at PNNL and the resources of the NERSC, a DOE SC user facility, using NERSC Award No. BES-ERCAP0035156 and BES-ERCAP-26557. This work also partially used Raven computational chemistry cluster at PNNL funded by the DOE under FWP 84864. We thank John L. Fulton, Christopher J. Mundy, Paul Rigor, Daniel Mejia-Rodriguez, Benjamin A. Legg, and Gregory K. Schenter for helpful discussions.