🔬 materials science

Cesium Clustering and Fluoroberyllate Network Disruption in FLiBe: A Total Scattering and Molecular Dynamics Study

This study combines total scattering experiments, empirical potential structure refinement, and neural network molecular dynamics simulations to reveal that adding 5 mol% CsF to FLiBe causes extensive cesium clustering and slight disruption of the fluoroberyllate network, which significantly suppresses the formation of the crystalline Li $_2$ BeF $_4$ phase at room temperature.

Original authors: Sean Fayfar, Rajni Chahal, Danny Wang, David J. Sprouster, Shravan Venugopal, Guiqiu Zheng, Dan Olds, Joerg Neuefeind, Stephen Lam, Boris Khaykovich

Published 2026-09-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Sean Fayfar, Rajni Chahal, Danny Wang, David J. Sprouster, Shravan Venugopal, Guiqiu Zheng, Dan Olds, Joerg Neuefeind, Stephen Lam, Boris Khaykovich

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Modern nuclear reactors are exploring a shift away from the water-cooled systems of the past, looking instead toward molten salts as a way to carry fuel and remove heat. These liquid salts, which remain stable at temperatures far higher than water can withstand, offer a path to more efficient and safer energy generation. However, as these reactors operate, they produce fission products—new atoms created when fuel breaks apart—that dissolve into the salt. One of the most significant of these byproducts is cesium. While scientists know that cesium accumulates in the salt, they have not fully understood how its presence changes the microscopic arrangement of the atoms within the liquid. Since the physical properties of the salt, such as how it flows or conducts heat, depend entirely on this atomic structure, knowing exactly how cesium rearranges the liquid is critical for predicting the long-term behavior of these future reactors.

To solve this puzzle, a team of researchers turned their attention to a specific molten salt mixture known as FLiBe, which is composed of lithium, beryllium, and fluorine. They added a small amount of cesium fluoride to this mixture, representing the concentration one might expect in a working reactor, and then subjected it to intense scrutiny. The team did not rely on a single method; instead, they combined high-energy X-ray and neutron diffraction measurements with advanced computer simulations. By firing beams of X-rays and neutrons at the hot liquid salt, they could see how the atoms scattered the particles, creating a map of the distances between them. They then used these experimental maps to refine computer models, ensuring the digital representations of the salt matched the real-world data as closely as possible. This dual approach allowed them to distinguish between features that were clearly visible in the data and those that remained dependent on the assumptions of the computer models.

The researchers found that adding cesium did not tear the liquid salt apart. The core structure of the FLiBe mixture, which consists of beryllium and fluorine atoms linked together in a network of tetrahedral shapes, remained largely intact. The cesium atoms did not force the beryllium and fluorine into a new, chaotic arrangement. Instead, the cesium ions behaved in a way that was somewhat unexpected: they tended to group together in clusters. Rather than spreading out evenly throughout the liquid, the cesium atoms gathered in specific regions, with the beryllium-fluorine networks acting as bridges that connected these cesium groups. This clustering was a dominant feature, visible in both the experimental data and the computer simulations, suggesting that cesium has a strong preference for being near other cesium atoms even within a liquid environment.

While the overall network held together, the presence of cesium did cause subtle shifts in the local environment. The addition of cesium slightly disrupted the connections between the beryllium and fluorine units, leading to a small increase in the number of free fluorine ions that were not part of the main network. The researchers also observed that the cesium atoms were surrounded by a specific number of fluorine neighbors, forming a coordination shell that was distinct from the rest of the liquid. Interestingly, the way the cesium clustered was more pronounced in the computer simulations than in the experimental data, indicating that while the general trend of clustering is real, the exact size and extent of these groups might be sensitive to the details of the model used. This distinction is vital, as it highlights which aspects of the salt's structure are firmly established by measurement and which require further refinement in theoretical models.

The study also revealed a stark difference between how the salt behaves as a liquid versus how it behaves as a solid. When the researchers cooled the mixture containing cesium, the salt refused to form the crystalline structure that pure FLiBe typically adopts at room temperature. Instead of solidifying into a neat, ordered lattice immediately upon cooling, the crystalline Li2BeF4 phase did not appear until the mixture was heated above 180 degrees Celsius. This suggests that the cesium acts as a powerful inhibitor of crystallization, preventing the salt from forming its typical solid structure at temperatures where it would normally do so, and delaying the appearance of the ordered phase until the mixture is heated well above room temperature. This finding is crucial for reactor safety and operation, as it implies that the presence of fission products could significantly alter the melting and freezing points of the fuel salt, potentially affecting how the reactor is started up or shut down.

Ultimately, this work provides a clear, experimentally grounded picture of how a common nuclear byproduct interacts with molten salt fuel. It confirms that while cesium does not destroy the fundamental architecture of the liquid, it does rearrange the local landscape by clustering and slightly loosening the network of connections. By combining direct observation with sophisticated modeling, the researchers have established a benchmark for future studies, ensuring that predictions about reactor performance are based on a structure that has been verified against reality. This level of detail is essential for engineers designing the next generation of nuclear power, as it allows them to anticipate how the fuel will evolve over time and how the physical properties of the salt will change as it accumulates the products of the nuclear reaction.

Technical Summary: Cesium Clustering and Fluoroberyllate Network Disruption in FLiBe

Problem Statement
Next-generation fission reactor designs, particularly molten salt reactors (MSRs), utilize FLiBe (a 2:1 eutectic mixture of LiF–BeF $_2$ ) as a coolant and fuel carrier. During operation, fission products such as cesium accumulate within the salt, potentially altering its thermophysical properties. While the atomic structure of pure FLiBe is relatively well-characterized, the structural perturbations induced by dissolved fission products like cesium (specifically as CsF) remain poorly understood. Accurate prediction of salt property evolution requires a precise understanding of how cesium interacts with the fluoroberyllate network at the atomic level, specifically regarding coordination environments, network connectivity, and clustering behavior.

Methodology
The study employs a multi-faceted approach combining experimental diffraction with advanced computational modeling to resolve the structure of FLiBe containing 5 mol% CsF:

Experimental Measurements: X-ray diffraction (XRD) and neutron diffraction (ND) were performed on FLiBe–CsF samples across a temperature range from room temperature to 700°C. XRD data were collected at the National Synchrotron Light Source II (NSLS-II), and ND data at the Spallation Neutron Source (SNS).
Empirical Potential Structure Refinement (EPSR): Experimental data were used to refine EPSR models. This technique bridges conventional atomistic simulations and Reverse Monte Carlo (RMC) approaches, incorporating interatomic potentials and experimental constraints to derive atomic configurations that satisfy both physical potentials and scattering data.
Neural Network Molecular Dynamics (NNMD): To complement the experimental constraints, the authors trained a DeepPot-SE neural network interatomic potential (NNIP) using Ab-Initio Molecular Dynamics (AIMD) data generated via Density Functional Theory (DFT). This NNIP was subsequently used to perform large-scale molecular dynamics simulations (2,565 atoms) at 510°C and 700°C.
Structural Analysis: Coordination numbers (CN) were calculated using two distinct methods: the traditional Geometric Cutoff (GC) method and the Relative Angular Distance (RAD) method. The RAD method was utilized to more accurately distinguish between first and second coordination shells, particularly for like-charge pairs and complex environments where intervening atoms might block direct coordination.

Key Results

Solid-State Behavior: Upon cooling, the FLiBe–CsF mixture did not form the expected crystalline Li $_2$ BeF $_4$ phase at room temperature; instead, it exhibited a complex mixture of unidentified Cs-containing crystalline phases. The Li $_2$ BeF $_4$ phase only emerged upon heating above 180°C, suggesting that the addition of CsF suppresses the formation of the standard FLiBe crystal structure at room temperature.
Liquid Structure and Total Scattering: The addition of 5 mol% CsF caused significant changes in the total structure factor and pair distribution function (PDF), primarily driven by new Cs–F and Cs–Cs correlations. However, the underlying FLiBe correlations (Be–F, Li–F, F–F) remained minimally altered.
Coordination Environments:
- Be–F Network: The local coordination of the fluoroberyllate network remained largely intact, with Be atoms predominantly maintaining fourfold coordination (BeF $_4^{2-}$ tetrahedra).
- Cs–F Coordination: Cesium ions exhibited a broad coordination distribution. Using the RAD method, the average Cs–F coordination number was found to be approximately 5.3, significantly lower than the ~10.0 value obtained via the GC method. This indicates that many atoms counted in the GC shell are actually in the second shell, blocked by intervening Be or Li atoms.
- Network Disruption: Cesium slightly disrupted the intermediate-range fluoroberyllate network. Both NNMD and EPSR models indicated an increase in the fraction of "free" fluorine ions (those not participating in the BeF $_4^{2-}$ network) and a shift from polymeric species toward monomeric BeF $_4^{2-}$ units.
Cesium Clustering: A significant finding is the extensive clustering of cesium ions within the molten salt. These clusters are not metallic but are mediated by BeF $_4^{2-}$ tetrahedra bridging neighboring cesium environments.
Model Comparison:
- EPSR vs. NNMD: EPSR models, constrained by experimental scattering data, accurately captured the prominent Cs–F and Cs–Cs correlations and the resulting clustering. However, EPSR models predicted longer polymer chains and more linear Be–F–Be bond angles, likely due to the neglect of polarization effects.
- NNMD Accuracy: NNMD simulations, trained on DFT data, successfully captured polarization effects, resulting in a more realistic Be–F–Be bond angle distribution (peak at ~130°) and a more accurate prediction of network polymerization. However, NNMD simulations predicted more extreme cesium clustering than EPSR, potentially due to the small number of Cs atoms in the training set.

Significance and Claims
The authors claim that this study provides a critical experimental benchmark for atomistic models used to predict the behavior of fission-product-containing FLiBe. By comparing EPSR (experimentally constrained) and NNMD (physics-based) results, the study successfully distinguishes between structural features that are directly resolved by scattering data (such as Cs–F and Cs–Cs correlations) and those that remain model-dependent (such as the precise extent of fluoroberyllate network polymerization and Be–F–Be angles).

The work demonstrates that while cesium induces only minor local perturbations to the FLiBe structure, it significantly alters the intermediate-range order by disrupting the network connectivity and promoting extensive cesium clustering. These experimentally constrained structural features serve to validate and refine simulation models, thereby improving confidence in the prediction of thermophysical property evolution in reactor-relevant conditions. The study does not propose new applications but rather establishes a rigorous methodology for characterizing complex, multi-component molten salts under reactor operating conditions.

More like this