Seven fragments cut from the AM (U-9% Mo)+Mg fuel rods, each 245 mm long and with a total mass of 1.565 kg, were put in a leak-tight canister underwater. Before that, the oxide layer was removed from both ends of the fuel fragments by cutting off 1 mm. Argon filled up the free volume of the canister at a pressure of 1 bar. The fuel rod fragments remained inside the canister for 380 hours, with the gas samples taken periodically. In 8 days after the soaking started, the water was sampled for the gamma spectrometric analysis that revealed ¹³⁷Cs with the total activity of (1.6 ± 0.2)×10⁸ Bq in 2.9 l of the solution.

The characterization of the fuel rods before the experiment determined that the maximum linear activity of ¹³⁷Cs in the fuel rods was 9.74·10⁹ Bq/cm. According to the findings, the depth of ¹³⁷Cs leaching on each end of the fragment was 12 µm roughly as of the time of sampling.

The gas sample from the canister contained hydrogen only. Fig. 1 demonstrates the kinetics of hydrogen generation approximated best by an exponential function that formally corresponds to the chemical first-order reaction

V = 356×[1 – exp(–0.0075t)], (1)

where V is the volume of hydrogen released under normal conditions, cm³; t is the time, h.

The contribution of the radiolytic hydrogen released from the SNF canister during the underwater storage was assessed based on the maximum specific activity of ¹³⁷Cs in the (U-9%Mo)+Mg fuel.

The accumulation rate for radiolytic hydrogen was calculated by the equation:

v = V_M×G_g×D_Cs×m_w/(100×N_A×1.6×10^–19×3600), (2)

where v is the accumulation rate for radiolytic hydrogen under normal conditions, l/s; V_M is the molar volume of the gas under normal conditions (22.4 l/mol); Gγ is the radiolytic G-value for molecular hydrogen per 100 eV absorbed (0.45 molecules/100 eV (Kabakchi and Pikayev 1981, Bugayenko et al. 1988)); D_Cs is ¹³⁷Cs dose rate, Gy/h; m_B is the mass of water in the canister, g; N_A is Avogadro’s number (6.02×10²³ molecules/mol).

The estimates did not take into account the radiolysis of water due to α- and β-emissions, since these particles spread only from the thin layer at the fuel ends as thick as the particle path in uranium metal (17.5 mg/cm² (Available at: ASTAR (NIST 2020a)) or 9 µm for 5 MeV α-particles with a density of 348 mg/cm² (Available at: ESTAR (NIST 2020b)) or 183 µm for 500 keV β-particles). The total activity of β-emitting radionuclides in the AM reactor fuel is one to two orders of magnitude less than that of ¹³⁷Cs; the total activity of β-emitting radionuclides can be taken equal to that of ¹³⁷Cs.

The absorbed γ-radiation dose rate was calculated with the MAVRIC functional module (Peplow 2009) of SCALE6.1.2 code package (ORNL/TM-2005/39 2011) developed by Oak Ridge National Laboratory for the U.S. Nuclear Regulatory Commission. The MAVRIC module of the SCALE code package calculates the particle intensity and the dose rate from a source of complex geometry based on Monte Carlo methods.

The absorbed dose rate in the water inside the canister was estimated conservatively for different arrangements of the fuel rod fragments in the canister. One of the fragments was placed along the central axis of the canister, while the others formed a peripheral row being placed symmetrically at the same distance from the axis.

Fig. 2 demonstrates a computational model of the canister containing seven fragments of AM fuel rods.

Fig. 3 presents the calculated absorbed dose rate in the water of the canister as a function of the circle radii for the peripheral fuel rod fragments. It identified the geometry with the maximum dose rate (12.9 Gy/h) for the water in the canister. The estimated accumulation rate for radiolytic hydrogen under normal conditions was 1.1·10^–8 l/s. It corresponds to 15 cm³ of hydrogen accumulated during the experiment, i.e. 4.5% of the measured value. Thus, it is the electrochemical corrosion of the fuel components that primarily governs the kinetics of hydrogen accumulation during storage of the AM (U-9%Mo)+Mg spent fuel.

Structurally, the AM reactor fuel rods represent particles of the U-9%Mo alloy dispersed in the Mg matrix between two cladding tubes made of 06Х18Н10Т stainless steel (Fig. 1). It causes galvanic couples between dissimilar metals and intense electrochemical corrosion when the fuel rod is in water. In this system of metals, magnesium (E_Mg+2/Mg = – 2.363 V) and uranium (E_U+3/U = – 1.789 V, E_U+4/U = – 1.5 V) have the most negative standard electrode potentials as compared to molybdenum and stainless steel (Golosov et al. 2013). So, the corrosion processes involve mainly the generation of insoluble hydroxides by the following reactions with hydrogen depolarization:

Mg + 2H₂O = Mg(OH)₂↓ + H₂, (3)

U + 4H₂O = UO₃×H₂O↓ + 3H₂, (4)

U + 5H₂O = UO₂(OH)₂×H₂O↓ + 3H₂, (5)

U + 4H₂O = UO₂(OH)₂ ↓ + 3H₂. (6)

The lower hydrogen release observed during the experiment can be an attribute of the insoluble corrosion products that impede access to the metal surface for the water, thus gradually decreasing the rate of reactions (3) – (6) until they cease.

Assuming that corrosion on the exposed surface (~12.6 cm²) induces the hydrogen accumulation, we can conclude that the corrected specific release of hydrogen per unit area of the exposed surface will not exceed 30 ml/cm² for any period of the underwater storage of the fragmented AM spent fuel with the magnesium matrix. On the assumption that all the hydrogen was generated by reaction (3), the estimated depth of the fuel layer on both ends of the specimen that had reacted with the magnesium matrix was 370 µm. It is much more than that estimated by the depth of ¹³⁷Cs leaching from the spent fuel.

Similar laboratory-scale corrosion tests of the AMB (U-3% Mo)+Mg and (U-9% Mo)+Mg spent fuel specimens are described in (Golosov et al. 2013). With the (U-9% Mo)+Mg specimen stored in a 800 cm³ canister at +40 °С for 270 days under a layer of water (60 cm³) in contact with air, the hydrogen release rate did not slow down, and oxygen escaped from the free volume of the canister. The spent fuel specimens were the fuel fragments machined (“coolers”) to include both the cladding and the fuel with annular notches on the external cladding and the fuel column to enlarge the total surface of the fuel column (about 40 cm²). Rough estimates in paper (Golosov et al. 2013) demonstrate that the amount of hydrogen accumulated per unit area for the fuel stored in canisters underwater for 270 days with air in the free volume of the canisters is an order of magnitude less than for the AM fuel stored in canisters underwater for 16 days with an inert gas in the free volume. It can be due to oxygen depolarization in the presence of oxygen in the air

Mg + 0.5O₂ + H₂O = Mg(OH)₂↓, (7)

U + 1.5O₂ + H₂O = UO₃×H₂O↓. (8)

But the experiment with the fragmented (U-9% Mo)+Mg fuel rods from the AM reactor in the absence of oxygen in the free volume of the canister demonstrated an evident effect of chemical passivation on the fuel surface during reactions (3) – (6) with hydrogen depolarization due to insoluble corrosion products generated.

The experiment used a UC+Ca fuel rod fragment 10 mm long with a mass of 7.12 g. The canister contained 0.98 l of water, with 2.6 l left for the free volume. The total time of keeping in water was 150 hours. The gas samples contained hydrogen and methane. Fig. 4 presents the kinetics of their release. An incubation period of about 20 hours was followed by an intensive gas release. The gas release rate decreased exponentially over time and nearly stopped in about 100 hours due to the total corrosion of the fuel composition.

During the experiment, the contribution of the radiolytic hydrogen released from the spent fuel canister was assessed based on the maximum specific activity of ¹³⁷Cs in the (U-9%Mo)+Mg fuel.

Assuming that the specimen had completely dissolved in the water and the radionuclides were evenly distributed in the solution, the hydrogen accumulation rate was calculated by the equation:

$v=\frac{V_M·G_{\gamma }·D_{Cs}·m_w}{100·N_A·1.6·{10}^{-19}·3600}+\frac{V_M·\left(G_{\alpha }·A_{\alpha }·E_{\alpha }+G_{\beta }·A_{\beta }·E·\beta \right)}{100·N_A}$, (9)

where G_α and G_β are the radiolytic G-values for molecular hydrogen per 100 eV absorbed α- and β-emitting energy (1.8 (Allen 1964, Vladimirova 1964) and 0.45 molecules/100 eV, respectively (Kabakchi and Pikayev 1981, Bugayenko et al. 1988)); A_α and A_β are the activities of α- and β-emitting radionuclides in the water inside the canister, Bq; E_α and E_β are the maximum energies of α- and β-emitting radionuclides, keV (5.5 and 0.5 MeV, respectively).

The γ-radiation dose rate was calculated with the MAVRIC functional module of SCALE6.1.2 code package. For α- and β-emitters, it was assumed that their energy was completely absorbed by the water in the canister. The activity of the β-emitting radionuclides was taken equal to that of ¹³⁷Cs. The α-activity of the specimen was unknown; however, it was conservatively taken equal to half the activity of ¹³⁷Cs based on the data that the specific α-activity of similar AMB spent fuel is 10–50 times lower than that of ¹³⁷Cs.

The estimated rate of the radiolytic hydrogen release was 3.8·10^-9 l/s. It accumulated only 2 cm³ of hydrogen during the entire experiment, i.e. 5% of the measured value.

The estimates, as for the Mg-matrix fuel, demonstrated that the hydrogen generation due to the water radiolysis is negligible.

The incubation period may be due to the slow dissolution of the corrosion products (oxides) generated on the fuel end cuts in air.

Based on the fuel assembly specifications sheet and the specimen geometry, uranium carbide and calcium masses were calculated to be 4.63 g and 0.88 g, respectively. When in contact with water, calcium generates calcium hydroxide and hydrogen as much as 560 cm³ per 1 g of Ca, while hydrolysis of uranium carbide yields hydrated uranium dioxide, the methane-hydrogen mixture, and a small amount of higher hydrocarbons (Bradley and Ferris 1961, Bradley et al. 1964, Hori and Mukaibo 1967, Peterson and Wymer 1967, Dyck et al. 1975).

Table 2 presents published data on the amount of the gases released per 1 g of uranium carbide and fractions of methane and hydrogen in the gas during the hydrolysis of irradiated and non-irradiated uranium carbide. The Table also provides the maximum estimated amounts of hydrogen and methane for the completely corroded experimental AM SNF specimen. The maximum possible hydrogen release due to the interaction of calcium with water is 493 ml for the experimental specimen under normal conditions.

Although the experiment with the AM spent fuel demonstrated that hydrogen was due to the UC hydrolysis and the calcium-water interaction, the estimates assume that only 25–40% of uranium carbide and 80–85% of calcium reacted with water in soaking the experimental UC + Ca fuel specimen.

According to (Bradley et al. 1964) and (Dyck et al. 1975), the higher is the burnup, the lower is the fraction of methane resulting from the UC hydrolysis, and the higher is the fraction of hydrogen. However, with the bunup higher than 4300 MW·day/tU, the specimens become non-reactive to water. Paper (Bradley et al. 1964) states that uranium carbide irradiated up to 6000 MW·day/tU does not undergo hydrolysis (no gas release) at all when kept underwater for 24 hours at least. Paper (Dyck et al. 1975) describes that the gas release was so low and the hydrolysis was so slow for the burnup of 4333 MW·day/tU and higher that the experiments were stopped before the reaction had ceased, although all the specimens had turned into powder in a short time.

The tested AM fuel specimen had high burnup (about 6270 MW·day/tU), but it witnessed a significant methane release after the incubation period of about 20 hours. This contrasts with the conclusions of papers (Bradley et al. 1964, Dyck et al. 1975) that uranium carbide irradiated up to a high burnup reacts with water poorly.