---
title: 'Uranium Mononitride (UN): Nuclear Fuel Material'
url: https://www.emergentmind.com/topics/uranium-mononitride-un
type: topic
---

# Uranium Mononitride (UN): Nuclear Fuel Material

Uranium mononitride, UN, is a binary uranium nitride that crystallizes in the NaCl-type rocksalt structure and has long been studied as a nuclear-fuel material with an unusual combination of metallic transport, antiferromagnetic order, partially itinerant \(5f\) electrons, and pronounced chemical reactivity toward oxygen and water. It has been treated as a promising candidate for Generation-IV and accident-tolerant fuel concepts because of its high uranium density, high thermal conductivity, and high melting point, but its deployment has been constrained by rapid oxidation, corrosion in aqueous environments, and the complexity of its correlated \(5f\)-electron physics [1005.5510] [1810.12156] [2102.07512].

## 1. Crystal chemistry and foundational materials properties

UN is described as a rock-salt structured, metallic compound and as an antiferromagnet below a Néel temperature of about \(T_N \approx 53\ \mathrm{K}\). In the ordered state it is type-I antiferromagnetic, whereas above \(T_N\) it is paramagnetic and cubic. Room-temperature lattice parameters reported for bulk or near-bulk material cluster near \(4.89\ \text{\AA}\), including \(a_0 = 4.888\ \text{\AA}\) in experiment, a room-temperature lattice parameter near \(4.89\ \text{\AA}\), and epitaxial-film values of \(4.895 \pm 0.001\ \text{\AA}\) [1005.5510] [1907.11685] [1803.02226] [2504.19184].

First-principles work has emphasized that standard LDA or GGA is insufficient for UN because it underestimates the on-site Coulomb repulsion of the U \(5f\) shell and incorrectly favors a ferromagnetic metallic ground state. Within GGA+\(U\), a Hubbard \(U\) around \(2\ \mathrm{eV}\) restores the experimentally correct AFM ground state and yields \(a_0 \approx 4.896\text{–}4.926\ \text{\AA}\) and \(B \approx 194.5\ \mathrm{GPa}\), close to experimental bulk moduli of about \(194\), \(200\), or \(206\ \mathrm{GPa}\). Bader analysis in the same framework gives an effective ionic picture \(\mathrm{U}^{1.71+}\mathrm{N}^{1.71-}\), while calculated Born effective charges are \(Z^*_{\mathrm{U}}=+1.95\) and \(Z^*_{\mathrm{N}}=-1.95\) [1005.5510].

The phonon spectrum contains six branches, shows no imaginary phonon frequencies, and reproduces the transverse optical mode at \(\Gamma\) as \(12.19\ \mathrm{THz}\), in excellent agreement with the experimental \(12.3\ \mathrm{THz}\) at \(4.2\ \mathrm{K}\). The phonon density of states separates into a uranium-dominated \(0\text{–}4.7\ \mathrm{THz}\) region and a nitrogen-dominated \(10.5\text{–}15\ \mathrm{THz}\) region, reflecting the large mass difference between U and N [1005.5510].

These results establish UN as a metallic actinide nitride with appreciable ionicity, stable lattice dynamics, and a magnetic ground state that is sensitive to correlation treatment.

## 2. Electronic structure, \(5f\) correlations, and magnetic-field response

Soft-x-ray ARPES has shown that UN possesses strongly dispersive bands with large contributions from U \(5f\) states and highly three-dimensional Fermi surfaces. In the paramagnetic phase, both the band structure and the Fermi surfaces are well explained by an LDA calculation that treats all U \(5f\) electrons as itinerant. The antiferromagnetic transition produces only very small changes in the spectral function, and the measured Fermi-surface geometry makes nesting an unlikely origin of the magnetic ordering [1211.4679].

Other work has argued that UN cannot be described adequately by a purely localized or purely itinerant picture. High-field magnetotransport, Hall-effect, and magnetic-torque measurements on microstructured single crystals identify a metamagnetic transition near \(58\text{–}60\ \mathrm{T}\) for \(\mu_0 H \parallel [100]\), evolving from a broad anomaly at higher temperature into a sharp first-order-like step below about \(10\ \mathrm{K}\), consistent with a spin-flop transition. Above the transition, the magnetic contribution to resistivity is strongly suppressed and the low-temperature transport becomes approximately quadratic, \(\rho(T)\propto T^2\), while the Hall effect shows a prominent anomaly interpreted as evidence for a substantial Fermi-surface reconstruction [2102.07512].

A more recent DFT+DMFT analysis with cRPA-derived interaction parameters \(U_C = 1.87\ \mathrm{eV}\) and \(J_H = 0.59\ \mathrm{eV}\) has identified UN as a strongly correlated bad metal in which the dominant correlations arise from intra-atomic Hund’s exchange between two \(5f\) electrons rather than from proximity to a Mott transition or from conventional Kondo-lattice physics. In that picture, the \(j=5/2\) states form a sharp peak pinned near the Fermi level, the \(j=7/2\) states appear as a broad incoherent feature centered about \(1\ \mathrm{eV}\) above \(E_F\), the U-\(5f\) occupancy is \(2.40\) electrons with SOC, and the most correlated \(j=5/2\) components have \(Z \approx 0.57\text{–}0.59\) at \(100\ \mathrm{K}\), implying \(m^\star/m \sim 1.75\) [2509.23033].

Taken together, these results suggest a material in which itinerant Fermi-surface physics, magnetic order, and strong local \(5f\) correlations are all simultaneously relevant. That synthesis is close to the “dual localized/itinerant” description advanced in the high-field study, even though individual probes emphasize different limits of the electronic structure [2102.07512].

## 3. Surface chemistry, oxidation, and grain-boundary reactivity

The most persistent materials limitation of UN is its high reactivity in oxygen-containing environments. On the defect-free UN(001) surface, DFT-GGA calculations show that \(\mathrm{O_2}\) molecules oriented parallel to the surface can dissociate spontaneously when the molecular center lies above a hollow site or atop a surface N ion, whereas dissociation from atop a surface U ion requires an activation barrier. The same study found strong charge transfer to the adsorbate, with roughly \(1\) electron transferred per \(\mathrm{O_2}\) molecule, and interpreted the atop-U geometry as a metastable UO\(_2\)-like quasi-molecular state [1008.5323].

Subsequent calculations on UN(001) and UN(110) extended this picture from adsorption to migration and vacancy capture. Oxygen adsorption is much more favorable atop U than atop N, stronger by about \(1.9\ \mathrm{eV}\) on UN(001) and \(2.2\ \mathrm{eV}\) on UN(110), with effective charges on adsorbed O of about \(-1.09\) to \(-1.10\ e\). On UN(001), oxygen adatoms are highly mobile: the migration barrier between nearest-neighbor surface U atoms is \(0.35\ \mathrm{eV}\) for a 5-layer slab and \(0.26\ \mathrm{eV}\) for a 7-layer slab. Oxygen penetration into a neighboring N vacancy requires only about \(0.5\ \mathrm{eV}\), and the overall oxygen solution energy is predicted to be close to about \(-2.5\ \mathrm{eV}\) for surfaces and grain boundaries alike [1211.5695].

The decisive defect-selective step is oxygen incorporation into nitrogen vacancies rather than uranium vacancies. For UN(001), incorporation energies are strongly negative for O entering N vacancies in the surface, subsurface, and central layers, whereas incorporation into U vacancies is near zero or positive. The electronic structure remains conducting, with strong U \(5f\) weight at the Fermi level and a distinct O \(2p\) band around \(-6\ \mathrm{eV}\); charge-density redistribution indicates local O–U bond formation around the occupied N vacancy [1008.5311].

The same chemistry extends to interfaces. In a \((310)[001](36.8^\circ)\) symmetric tilt grain boundary, the N-vacancy formation energy is \(3.34\text{–}3.48\ \mathrm{eV}\), compared with about \(4.3\text{–}4.6\ \mathrm{eV}\) in bulk-like central layers, while oxygen incorporation into a pre-existing N vacancy is strongly favorable at \(-5.67\) to \(-5.92\ \mathrm{eV}\) and the corresponding oxygen solution energy is \(-2.33\) to \(-2.45\ \mathrm{eV}\). Grain boundaries therefore behave much like free surfaces in attracting N vacancies and trapping oxygen [1211.5693].

A compact thermodynamic description was written in terms of the Gibbs free energies of oxygen incorporation and N-vacancy formation,
\[
\Delta G_O(T) = -\left(E_{UN+O}-E_{UN-vac}-\mu_O(T)\right),
\]
\[
\Delta G_F(T) = -\left(E_{N_v}-E_{UN}+\mu_N(T)\right),
\]
with temperature-dependent chemical potentials for O and N [1211.5695].

This body of work supports an interface-controlled oxidation mechanism: \(\mathrm{O_2}\) adsorbs, dissociates, atomic O binds to surface U, migrates rapidly, and is incorporated into N vacancies at surfaces and grain boundaries, promoting formation of oxynitrides \(UO_xN_y\). Experimentally, the same study notes that UN begins oxidizing around \(250^\circ\mathrm{C}\), with oxide phases dominating by about \(500^\circ\mathrm{C}\) [1211.5695].

## 4. Thermodynamic behavior and thermal transport

Within GGA+\(U\) and the quasiharmonic approximation, the Helmholtz free energy of UN was written as
\[
F(V,T) = E(V) + F_{\mathrm{ph}}(V,T) + F_{\mathrm{ele}}(V,T),
\]
and the calculated specific heat was found to agree well with experiment up to about \(200\ \mathrm{K}\) without electronic excitation, while inclusion of a conduction-electron contribution improves agreement substantially at higher temperature. The calculated electronic specific-heat coefficient is
\[
\gamma = 26.7\ \mathrm{mJ\,K^{-2}\,mol^{-1}},
\]
smaller than the experimental \(49.6\ \mathrm{mJ\,K^{-2}\,mol^{-1}}\) but much larger than in ThN [1005.5510].

Because reactor-relevant conditions are paramagnetic, ab initio molecular dynamics combined with the disordered local moment approach has been used to model UN above \(T_N\). In that framework, the local U moments are randomized every \(t_{\mathrm{SF}}=5\ \mathrm{fs}\) during NPT AIMD in a \(3\times 3\times 3\) supercell. The resulting paramagnetic UN remains cubic from \(300\) to \(2000\ \mathrm{K}\), and the lattice parameter is fitted as
\[
a(T) = 4.862 + 1.593 \times 10^{-5} T + 8.719 \times 10^{-9} T^2,
\qquad R^2 = 99.6\%.
\]
The calculated \(C_P\) overestimates experiment at roughly \(300\text{–}500\ \mathrm{K}\), slightly underestimates it above \(500\ \mathrm{K}\), and nevertheless follows the experimental trend closely [2504.19184].

At still higher temperature, the intrinsic form of \(C_P(T)\) remains uncertain above roughly \(1700\ \mathrm{K}\). Large-scale MD with two interatomic potentials has examined whether nitrogen Frenkel-pair formation contributes to the observed superlinear rise. Using a defect-thermodynamics estimate
\[
C_{\mathrm{def}}(T) \approx \frac{Q}{n_{\mathrm{moles}}}\frac{\partial n_{\mathrm{FP}}}{\partial T},
\]
with \(Q = 4.5\ \mathrm{eV}\), the Tseplyaev potential yields a defect heat-capacity contribution of about \(10\ \mathrm{J/(mol\!-\!K)}\) at \(2600\ \mathrm{K}\), whereas the Kocevski potential gives only about \(1\ \mathrm{J/(mol\!-\!K)}\). This suggests that nitrogen sublattice disorder is a plausible intrinsic mechanism for the high-temperature curvature of \(C_P(T)\), though the quantitative magnitude remains model-dependent [2603.02441].

Thermal transport in UN is intrinsically mixed electronic and phononic:
\[
\kappa_{\mathrm{total}} = \kappa_{\mathrm{L}} + \kappa_{\mathrm{e}}.
\]
In pristine UN, MLIP-based Green–Kubo and ShengBTE calculations agree closely when both three-phonon and four-phonon processes are included, leading to the conclusion that four-phonon scattering is essential and sufficient to capture high-temperature lattice thermal transport over \(300\text{–}1500\ \mathrm{K}\). One study found that electronic contributions dominate thermal transport above \(600\ \mathrm{K}\), while another placed the crossover above about \(800\ \mathrm{K}\); both agree that \(\kappa_{\mathrm{e}}\) becomes increasingly important at elevated temperature [2605.17726] [2507.18786].

The same transport literature has also resolved specific scattering channels. In the MTP-based thermal-conductivity study, including four-phonon scattering reduces \(\kappa_L\) by about \(40\%\) at \(900\ \mathrm{K}\), and electron–phonon scattering suppresses \(\kappa_L\) by about \(80\%\) at \(77\ \mathrm{K}\), becoming negligible above about \(600\ \mathrm{K}\). Total thermal conductivity was reported to agree well with experiment from \(77\ \mathrm{K}\) to \(900\ \mathrm{K}\) [2507.18786].

## 5. Point defects, diffusion, and high-temperature deformation

Point defects play a central role in both heat transport and mechanical response. A recent defect-transport study considered uranium interstitials \(\mathrm{I_U}\), uranium vacancies \(\mathrm{V_U}\), nitrogen interstitials \(\mathrm{I_N}\), and nitrogen vacancies \(\mathrm{V_N}\), all at \(0.46\%\) concentration. All defect types reduce lattice thermal conductivity at low temperature. For lattice conductivity, the reduction severity follows
\[
\mathrm{I_U} > \mathrm{I_N} > \mathrm{V_U} > \mathrm{V_N},
\]
whereas total conductivity degradation, after inclusion of electron–defect scattering, follows
\[
\mathrm{I_U} > \mathrm{V_U} > \mathrm{I_N} > \mathrm{V_N}.
\]
Interstitials introduce new phonon states because of stronger local strain effects; \(\mathrm{I_U}\) causes broad defect-phonon scattering across the spectrum, and \(\mathrm{I_N}\) introduces localized gap states within the phonon bandgap [2605.17726].

Machine-learning potential studies of diffusion likewise indicate strong defect assistance. Nitrogen self-diffusion is difficult to resolve in stoichiometric UN except at higher temperatures, but nitrogen vacancies increase diffusivity and nitrogen interstitials increase it even more strongly, supporting the interpretation that anion diffusion in UN is mainly interstitial [2411.14608]. Large-scale MD for \(1800\text{–}2600\ \mathrm{K}\) further showed rising nitrogen diffusivity and Frenkel-pair populations, with the Tseplyaev potential giving \(D_{\mathrm N}\) increasing from about \(3.7\times10^{-15}\ \mathrm{m^2/s}\) at \(1900\ \mathrm{K}\) to \(3.4\times10^{-11}\ \mathrm{m^2/s}\) at \(2600\ \mathrm{K}\) [2603.02441].

High-temperature deformation has been analyzed with nanocrystalline MD. For polycrystals with grain diameters of \(14\), \(16\), \(18\), and \(20\ \mathrm{nm}\), equilibrated at \(1700\text{–}2000\ \mathrm{K}\) and stressed at \(200\text{–}500\ \mathrm{MPa}\), Nabarro–Herring creep was found not to be dominant. Instead, Coble creep controls diffusional deformation, with a grain-boundary-diffusion activation energy
\[
Q_{\mathrm{eff}} = 2.28 \pm 0.09\ \mathrm{eV}
\]
and an effective grain-boundary width
\[
\delta = 2.69 \pm 0.08\ \mathrm{nm}.
\]
The Coble-creep law was written as
\[
\dot{\epsilon}_C=A_C\frac{\sigma \Omega}{k_B T}\frac{D_\mathrm{GB}\delta}{d^3},
\]
and the parameterized UN form as
\[
\dot{\epsilon}=582610.427\,\frac{\sigma}{T d^3}\exp\!\left(-\frac{2.28\ \text{eV}}{k_B T}\right),
\]
with a combined thermal-creep model obtained by adding the Hayes dislocation-creep correlation [2503.03231].

These results make UN a defect-sensitive material not only chemically but also thermally and mechanically: fine-grained microstructures, grain boundaries, and anion disorder all change transport and deformation in quantitatively important ways.

## 6. Epitaxial films, strain effects, and corrosion screening

The first single-crystal epitaxial thin films of UN and \(\alpha\)-U\(_2\)N\(_3\) were grown by reactive DC magnetron sputtering. In that work, both films were [001]-oriented and single-domain; the UN film had a specular lattice parameter of \(4.895 \pm 0.001\ \text{\AA}\), a miscut of \(2.6^\circ\), and an XPS-derived N/U ratio of about \(1.02 \pm 0.02\), consistent with stoichiometric UN within experimental error. \(\alpha\)-U\(_2\)N\(_3\) was highlighted as a key intermediate oxidation product in the pathway UN \(\rightarrow\) U\(_2\)N\(_3\) \(\rightarrow\) UO\(_2\) [1803.02226].

Synchrotron x-ray scattering on epitaxial UN revisited the long-standing magnetic-structure problem. The measurements confirmed antiferromagnetic order with \(q=\langle 001\rangle\), but did not detect the clear tetragonal distortion expected for a conventional single-\(k\) type-I AF state. Instead, the data placed an upper bound of \(|c/a-1|\lesssim 2.3\times10^{-4}\), found \(T_N = 45.8(3)\ \mathrm{K}\) in the strained film, and measured a magnetic critical exponent \(\beta = 0.53(5)\). The authors concluded that strain redistribution, not intrinsic tetragonal symmetry lowering, dominates the low-temperature structural response in the film and argued that the accepted AF configuration of UN may need re-evaluation, possibly toward a cubic \(3k\) state [1907.11685].

Thin films have also enabled direct corrosion screening. In \(0.1\ \mathrm{M}\ \mathrm{H_2O_2}\) at room temperature, used as a surrogate for radiolytic oxidants, x-ray reflectivity measurements gave corrosion rates of \(0.083(3)\ \text{\AA/s}\) for UO\(_2\), \(0.020(4)\ \text{\AA/s}\) for UN, and \(0.47(8)\ \text{\AA/s}\) for U\(_2\)N\(_3\). Under these conditions, UN corrodes more slowly than UO\(_2\), and the thickness of the oxidized surface layer on UN remained approximately constant with exposure time at around \(90\ \text{\AA}\), suggesting a possible rate-limiting step in UN oxidation [1810.12156].

That result does not overturn the well-established chemical reactivity of UN in water and air. The same corrosion study explicitly notes that \(0.1\ \mathrm{M}\ \mathrm{H_2O_2}\) is much higher than realistic radiolytic concentrations and should be interpreted as an initial radiolytic-corrosion screen rather than a full reactor-accident simulation [1810.12156].

## 7. Computational infrastructure and emerging modeling paradigms

UN has become a major benchmark system for advanced atomistic methods because it combines actinide \(5f\) electrons, defect-rich oxidation chemistry, and technologically important thermophysical behavior. Recent work reported the first machine-learning interatomic potentials for UN: an ANI-based neural network potential and a HIP-NN-based potential trained on a DFT dataset of \(12{,}336\) atomic configurations produced by active learning. At \(300\ \mathrm{K}\), these models gave lattice parameters of \(4.857\ \text{\AA}\) and \(4.858\ \text{\AA}\), close to the experimental \(4.89\ \text{\AA}\), and they reproduced heat capacity, bulk modulus, stoichiometric defect reactions, migration barriers, nitrogen diffusion trends, xenon incorporation energetics, and collision-cascade damage evolution [2411.14608].

A separate MTP study trained on \(3236\) DFT configurations reported validation errors of \(2.82\ \mathrm{meV/atom}\) for energy and \(45.49\ \mathrm{meV/\AA}\) for forces on a \(450\)-configuration validation set. The equilibrium lattice constant was \(4.87\ \text{\AA}\) in both MTP and DFT, compared with \(4.88\ \text{\AA}\) in experiment at \(300\ \mathrm{K}\), and the same potential predicted a melting point \(T_m \approx 3080\ \mathrm{K}\), close to the experimental recommendation \(3120 \pm 30\ \mathrm{K}\). Combined with a new low-temperature spatial-domain thermoreflectance measurement representative of single-crystal behavior, the study concluded that the developed potential reliably predicts thermal expansion, specific heat, and thermal conductivity [2507.18786].

Another line of work has attempted to push beyond the accuracy ceiling of DFT-pretrained models by refining them against EXAFS. For UN, a HIP-NN MLIP pretrained on a DFT dataset of \(12{,}336\) structures was refined at \(50\ \mathrm{K}\) using synthetic EXAFS targets for U and N. The refined potential improved the lattice parameter across \(300\text{–}2500\ \mathrm{K}\) and reduced the bulk modulus from \(219.48\ \mathrm{GPa}\) to \(203.06\ \mathrm{GPa}\), closer to the experimental \(206.7 \pm 0.8\ \mathrm{GPa}\), while preserving reasonable point-defect energetics [2506.10211].

The computational landscape of UN therefore spans several distinct but complementary levels: DFT+\(U\) for static structure and phonons, AIMD+DLM for paramagnetic thermophysics, DFT+DMFT for correlated-electron phenomenology, and MLIPs for mesoscale-relevant defects, diffusion, creep, and thermal transport. A plausible implication is that UN will continue to serve as a reference actinide nitride for integrating correlated-electron theory with fuel-performance modeling, particularly where surfaces, grain boundaries, and high-temperature disorder are all active simultaneously.

Source: https://www.emergentmind.com/topics/uranium-mononitride-un