Speaker: Anna Smith,TU Delft at GTT Users’ Meeting 2025, held on 4-6 June 2025 in Aachen, Germany
Abstract: From structure to thermodynamic and thermophysical properties of molten salt nuclear fuels: experimental and computational studies at the TU Delft
Anna L. Smith(1), Nick ter Veer(1), Lukasz Ruszczynski(1), J.A. Ocadiz-Flores(1), John Vlieland(1), Sebastian Couweleers(1), Aimen Gheribi(2), Kathy Dardenne(3), Joerg Rothe(3), Pier-Lorenzo Solari(4), Rudy J. M. Konings(1,5)
(1) Delft University of Technology, Radiation Science and Technology Department, The Netherlands
(2) Concordia University, Canada
(3) KARA-INE synchrotron beamline, Karlsruhe Institute of Technology, Germany
(4) SOLEIL-MARS synchrotron beamline, France
(5) Joint Research Centre-Karlsruhe, European Commission
Molten salts are receiving increasing attention worldwide as key materials for sustainable and low-carbon energy technologies, including for fission reactors, and in particular for the Molten Salt Reactor (MSR). Because of their appealing thermo-physical properties (e.g. low melting and high boiling points, low vapour pressure, thermal stability, high heat capacity and thermal conductivity etc.), molten salts are considered in MSRs both for the fuel and coolant. Fluoride and chloride salts are to this date the two main candidates, depending on the main objective pursued for the envisaged nuclear reactor, e.g. actinide burner, thorium breeder etc. One main challenge for their future commercialisation is a comprehensive understanding and modelling of the molten salt fuel chemistry, enabling a thorough safety assessment.
Because of the challenges inherent to the work with molten salt materials, which are hygroscopic, highly corrosive at high temperatures, and in the case of nuclear applications radioactive, the available knowledge on the physico-chemical properties is still far from complete. Among these, the local structure properties of the salts at high temperature are particularly relevant, as the formation of short-range order observed in the liquid, has a direct impact on the transport (e.g. viscosity) and excess thermodynamic properties.
In this presentation, I will discuss structural studies of the melt carried out in our research group, that are used as input to develop coupled models of the structural (i.e. chemical speciation) and thermodynamic properties of the molten fuel salt. I will also show examples of our latest efforts to develop models of the density and viscosity of the melt, coupled to the thermodynamic assessments. Developing sound models of the thermochemistry and thermophysical properties of the fuel salt will allow in fine to assess the effect on the fuel properties of e.g. fission and corrosion products accumulation during irradiation in the nuclear reactor.
Good morning everyone. My name is Anna Smith and I will tell you about our studies in Delft on the structure, thermodynamic and thermophysical properties of molten salt nuclear fuels. Nowadays, nuclear energy is generated around the world by lightwater reactors which are so-called generation 2 reactors which run on a solid fuel. It’s a uranium dioxide, the ceramic. And this fuel is cooled down by water, which operates at a temperature of around 2 320 to 350° C, meaning that the primary loop has to be pressurized. The energy from the nuclear reaction comes from the fishision of the uranium 235 which when it under goes fishision produces fish products, a few more neutrons and a lot of energy. And nowadays we are looking uh at the next generation of nuclear reactors that could replace the current fleet of reactors when they come at the end of a lifetime. We are investigating so-called generation 4 reactors and among these reactors, one of the design is called the molten salt reactor. It’s a very innovative design that is based not on a solid fuel but on a liquid fuel and coolant. These type of reactors are very interesting. They are very promising. They present some inherent passive safety features. They can operate at low pressures, atmospheric pressure. They have a very uh great flexibility for the fuel cycle and they show properties of good retention of fish products. I was talking about the fuel flexibility here in this type of reactors. In addition to the fishision of uranium 235, we can also run the reactor on plutonium 239 which could come from stock piles from civil nuclear energy or military stock piles. But it could also run on toum which is a fertile element that becomes file 200 uranium 233 after absorption of a neutron. So we do a lot of research on this type of reactors in Delft and to run this kind of reactor we need a liquid fuel with some specific properties. So we need to choose the fuel very carefully to satisfy some conditions on neutronic properties and things that I will talk about today uh related to chemical thermal and transport properties. We would like a fuel with a certain density, certain viscosity, certain thermal conductivity, but also a fuel with a high solubility of file elements. We want a fuel that has a low melting point, a high heat capacity, a low vapor pressure or a high boiling point. At this moment, uh two types of salts are being uh envisaged among the MSR community. uh some reactor designs are based on fluoride mixtures and other reactor designs are based on chloride mixtures. Usually there are two or three maybe four end members in these mixtures with a file element for example uranium tetra fluoride or plutonium trichloride uh an alkali a light for example here lithium fluoride but it could also be sodium fluoride potassium fluoride or if it’s a chloride sodium chloride mixture and maybe a fertile element if there is thorium in the mixture. So we are investigating this type of nuclear fuel and why is it so important? Uh why is nuclear fuel chemistry so important? This is because understanding it is key for the good performance of nuclear reactors. The chemistry of the fuel will have an impact on the behavior during normal operation but also during accidental uh scenario and accidental conditions. So really having a good insight on this fuel is key and at the heart of the progress of generation 4 technologies. Our approach in Delft is to perform some experimental measurements of phase diagram of thermodynamics of the systems. We also look at the structure of a salt and we couple this uh with molecular dynamic simulations to get more insight into the structure of the salt into thermophysical properties such as density and viscosity. And once we have this information from experimental measurements or from simulations, we use that as input to develop some thermodynamic models as well as density and viscosity models that are coupled to the thermodynamic assessments. We do that for several binary systems, turnary systems and in the end we want to build a large database with multiple components which will then allow us to perform application calculations to be able to assess uh safety scenarios as well as simply normal operating conditions. So today what I wanted to show you to build uh upon the previous presentation by Andre Benes from the GRC is uh our uh current effort to develop thermodynamic models that also describe the structure of a melt the structure of a nuclear fuel salt. To give you a bit of insight on the structure of molten salt, if uh we have purely sodium chloride, you could say table salt, we expect to have a strongly ionic liquid. But as soon as we add uh uranium triricchloride in mixture with sodium chloride for instance, we form a molecular type of liquid, we form monomeic species, complexes that can have various coordination. And these polyedua will start to form networks. They will start to form chains uh that will really become predominant when the concentration of uranium chloride in this case increases. And now if we had in the salt mixture something like berium fluoride then we could expect to form really a polymeric liquid at high burillium fluoride content in a mixture. we start to form dus and really polymeric chains when the concentration of radium fluoride increases. And why is this important? Uh this is because this structure of the salt has a direct impact on the thermophysical properties. If we take viscosity as an example, we see that the viscosity in a mixture of lithium fluoride with uranium tetra fluoride or berium fluoride increases as the content of uranium or berium fluoride increases. But there is quite a difference between the increase in lithium uranium fluoride mixture which is about one order of magnitude versus several orders of magnitude as you see here. If we add berium fluoride to the mixture. So we really want to understand this relationship between the structure of the salt, the thermodynamics and the thermophysical properties. How do we do that? Uh we use experiments uh uh we use X-ray absorption spectroscopy experiment which are performed at the Synotone facility and that’s because we need very intense and tunable X-ray beams. Since we also want to look at the molten salt, so a melt, we need to do high temperature in C2 experiments. So we have built in Delft uh a furnace setup to be able to heat up our molten salt systems at high temperatures and we bring this furnace to a senoton facility in Germany where we measure then uh the X-ray absorption coefficient uh of a certain sample. So when we shoot some X-rays uh to a sample, we can measure the intensity before the sample and after the sample and uh we in this way measure the X-ray absorption coefficient as a function of energy. What you see on this uh figure is an example of a measurement we did on a lithium fluoride to fluoride mixture at 920° C. And in this kind of measurements there are two regions of interest. The low energy region called the exanises and the high energy region called the exafs. It is this exaf that is mostly interesting in this case. It gives us some information on the local structure around uh in this example a toium absorbing atom. So we get information on the first coordination shell, the florine around the toum and maybe also the second coordination shell. Here is a picture our furnace set up at high temperature with some detectors in front. And so again this exaf region after some mathematical treatment uh of the data we get the exaf uh signal and after 4 year transform we get uh this uh uh figure which is a bit equivalent to pair distribution function where toum is here and we see here the first clo coordination shell of florine around vtorium absorbing atom. So when we fit this data we get some average information on the local structure we get information on the average distance between toium and florine in this case 2.3 angstrom as well as the average coordination number on average we found out that there was at this composition seven flowing ions around each toium. So that is already some information but not a very detailed information. If we want more detailed information, we need to couple this experimental approach with molecular dynamic simulations. We do that using a so-called polarizable ion model that is really well adapted to ionic systems. And with these MD uh potentials which we have, we can simulate exaf uh data and compare to our experiments. This allows us to validate the interaction potentials of the molecular dynamic simulations. And in case there is a discrepancy, we can also fine-tune the interaction potentials to get a better agreement with the experimental data. With that we have validated uh interaction potentials from molecular dynamic simulations and we can then use the MD to get a detailed uh information and insight on the structure at any temperature and composition uh in this system including in regions that we have not measured experimentally. We have followed this approach for example for a lithium fluoride uranium fluoride mixture and after the experimental campaign and after the molecular dynamic simulations we could see that uh the chemical speciation of the salt varies a lot with composition at low lithium fluoride content we have mostly monomeic species of uranium fluoride that are sevenfold 8fold and nf-fold coordinated and when we increase uranium fluoride content we start to form dus and eventually polymeric chain. So we have a fully connected network here. We then use this information to build our thermodynamic models. And how do we do that? We use the quasy chemical model approach which is also the one used in the GLC molten salt database that is really well adapted to ionic liquids that is based on a description of a liquid with quadruplets. We know however that uh this mathematical description of a liquid with quadruplets uh allows to reproduce very well the phase diagram data. Maybe also other properties such as mixing enthalpy but do not give a completely physical description of a melt where we have molecular species that form as well as an oligization of a melt as we have seen from exas and MD. So what we have uh done is to introduce a bit more structural complexity into this quasy chemical description by introducing some cations with a certain coordination. We knew for the lithium fluoride uranium fluoride system that we form at low lithium fluoride content mostly sevenfold and eight-fold coordinated uraniums. And when we increase the uranium fluorite concentration, we start to form dness. So we made uh a model of that system that included sevenfold coordinated uranium, eight-fold coordinated uranium as well as dus we could say that represents the formation of polymeric chains at high uranium content. In this way we can have a thermodynamic model that reproduces phase diagram information but also gets u some gives some insight on the structure of a liquid as we know it from the experiments and from the molecular dynamic simulations. So with this uh kind of description we can also reproduce the chemical speciation from the molecular dynamic simulation and this is what you see here where the line which is the kan model also reproduces the speciation of the monomeic species the sevenfold and eightfold coordinated uranium as well as the progressive formation of polymers in this melt when the uranium content increases. is so this is the first uh type of model I wanted to show today and the second one is our effort to build models that couple thermodynamics with density and viscosity since density and viscosity are also two very important properties for the operation of a molten salt reactor and for the safety assessment. What we would like is to use our thermodynamic models which use the quasi chemical formalism in the quadupplate approximation to build models of the nuclear fuel uh that can also describe density as well as viscosity in the coupled approach. And here you have an example of our work on uh binary systems of lithium fluoride with uranium and toium chloride as well as sodium chloride with uranium to chloride. So how do we do that? Uh the density is related to the molar mass and the molar volume and this molar volume can be uh expressed as follows. And this is um uh an expression that uh was uh developed uh in the previous work of Robl Charton already some years ago and so we have used the same approach and applied it to our nuclear fuels. So this molar volume is expressed as a stochometric sum of the molar volume of the n members plus an excess contribution which represents the deviation from ideality of this molar volume. So this uh excess molar volume is related to the quadruplet distribution. So this is related to thermodynamic model as well as to the partial derivative of the excess gibs energy with respect to pressure. And this term is really the core of the density model where we introduce some excess parameters the beta terms here which can be optimized and represent then this excess uh m volume or this deviation from ideality. Once we have done that, we can go one step further to develop a viscosity model where we have a target function the viscosity that is expressed with a naring type of equation. This one related to plank constant the avagadro number as well as a molar volume. So here is again a link with the density model and the thermodynamic model. And then we have this exponential term with this G star which is the molar activation energy for viscous flow expressed as a function of the quaduplate distribution and some terms that are optimized in the model. So here again you see that really we have a viscosity model that still is very much related to the thermodynamic description and to the description of density. We have done this uh as I mentioned for nuclear fuels. We have followed a kad type of approach where first we need to build models for the end members models for density and viscosity of uranium tetrachloride to tetraoride and uranium triricchloride. Once this is done, we can move on to binary mixtures where we describe uh the density in a binary mixture. And finally, we can go to higher order systems until we get description of density and viscosity for a turnary mixture. In this example, lithium fluoride, toium fluoride and uranium tetra fluoride. Finally uh moving on uh towards more complicated system. We have so far modeled fresh nuclear fuel but of course when fishision happens a lot of fish products are generated which can be metallic precipitates gas or salt soluble species and we would like to know how that uh fish will impact density and viscosity of the irradiated nuclear fuel. So we need to build in the end a very comprehensive database for the density and viscosity of multicomponent systems that can really represent nuclear fuel with fish products and that’s a work we are going to do within a European project called that just started about six months ago where we will build uh thermodynamic models density models viscosity models of systems of fish products uh so that we can build this large database for density and viscosity that will then be coupled to the GLC molten salt database presented uh just before. Finally, I want to acknowledge that this work uh was done together with uh my team in Delft. I have really wonderful co-workers. They’re all uh listed here. And of course, we also have a lot of collaborators. uh and that work would not be possible without the collaborators. So we collaborate with people at KIT ENA for the spectroscopy measurements. Our colleagues at the GRC culture at Concordia University Paris and Orano. Finally, I would like to thank you for your attention and let me know if you have questions.