Introduction

The need for a database characterizing the properties of selected materials to be used in future commercial fusion reactors has been generally recognized as one of the eight missions described by the European Fusion Roadmap [1]. This is the objective of the International Fusion Materials Irradiation Facility—DEMO-Oriented Neutron Source, or IFMIF-DONES, a research facility designed for testing, validating, and qualifying materials, which will produce a deuterium beam (40 MeV, 125 mA) impacting on a liquid lithium (Li) target [2]. The nuclear stripping reactions between Li and deuterons will generate high energy neutron fluxes that will irradiate the material samples in dedicated test modules, located immediately behind the Li target or jet.

In IFMIF-DONES, five major groups of systems can be identified: Site, Building, and Plant Systems; Accelerator Systems; Lithium Systems (LS); Test Systems; and Central Instrumentation and Control Systems. Under the Spanish regulations, the IFMIF-DONES facility will be a first-class radioactive facility, which means that the radiological and non-radiological hazards need to be evaluated in order to demonstrate an acceptable low-level risk for the public, workers, and environment. To analyse the safety risks, different tools are available, such as Failure Mode and Effect Analyses (FMEA) [3] and Probabilistic and Deterministic Analyses. These tools contribute to the identification of a list of Reference Accident Scenarios (RAS) and of those Safety Class Structure, Systems, and Components (SCSs) to be implemented to the IFMIF-DONES final design.

In this work, two potential scenarios related to the primary loop (PLO) of the Lithium Systems have been analysed with the MELCOR v1.8.6 for fusion code. For this purpose, a break in the inlet nozzle of the Target Vacuum Chamber (TVC) and a break at the bottom of the Quench Tank (QT) have been postulated as initiating events, leading in both cases to a Li spill into the Test Cell (TC) room. By means of the deterministic analysis performed on these scenarios, key metrics are obtained, including the mass of Li spilled in the room, the time it takes for the spilled Li to completely solidify, and the maximum temperature and pressure loads reached in the Test Cell. These parameters directly influence the final design of the Test Cell liner, and can help identifying the optimal design strategies for preventing and mitigating this type of accident scenarios which could potentially involve Li fires and the consequent mobilisation of toxic aerosols and radionuclides (e.g., tritium).

The Design of IFMIF-DONES Lithium Systems

In IFMIF-DONES, the Lithium Systems (LS) are composed by the following systems as shown in Fig. 1:

  • The Impurity Control System (ICS), which purifies the Li during normal operation,

  • The Heat Removal System (HRS), whose main functions are to provide the required lithium flow and to remove the thermal power deposited by the deuteron beam on the lithium target, and.

  • The Target System (TSY).

Fig. 1
figure 1

Basic configuration of the Lithium Systems [4]

Full size image

Specifically, the analyses developed in this work have focused on the PLO which is part of the HRS. The PLO includes an electromagnetic pump (EMP) and a primary heat exchanger (HX), and is directly connected to the TSY [4]. The TSY is located inside the TC room and its function is to create and steadily maintain the liquid Li jet with the required characteristics during the normal operation of the facility. Figure 2 shows the main components of the TSY, including the Flow Straightener and Reducer Nozzle, the Target Vacuum Chamber (TVC), the Quench Tank (QT) and, finally, the outlet/inlet Li pipes.

The EMP drives the liquid Li through the PLO, reaching the HX where the thermal power, previously deposited on the Li target by the deuterium beam inside the TVC, is extracted. Li continues to flow through the Li inlet pipe, reaching the Target Assembly (TA), generating the liquid Li curtain inside the TVC, then, it descends through the QT, reducing its speed, and, finally, returns to the EMP through the outlet pipe.

The floor of the TC will be cooled by a dedicated circuit to remove nuclear heating resulting from deuteron beam-on conditions on the target. Since this circuit is not considered a safety-credited system according to the current design, it has not been included in the computational model; thus, the analysis provides more conservative results regarding the temperatures achieved by the liner.

Fig. 2
figure 2

DONES target system [4]

Full size image

MELCOR v1.8.6 Code for Fusion

MELCOR is a system-level thermal-hydraulic code used for performing deterministic analyses that allow analysing the steady-state and transient behaviour of nuclear and other facilities during normal operation and accident scenarios [5]. It has been originally conceived by Sandia National Laboratories (SNL), and its most recent fusion version (v1.8.6) is being developed by the Idaho National Laboratory (INL). A previous version of MELCOR fusion has been used in support of ITER licensing process and its last version is currently being used for DEMO safety analysis [6, 7].

MELCOR allows performing thermal-hydraulic transient calculations, solving the equations of conservation of mass, energy, and momentum. The fusion version of the code allows the use of liquid metals (such as Li) and includes models for simulating Li fires [8] and aerosol transport [9].

Previous Deterministic Analyses in IFMIF-DONES

Within the safety activities being developed in the frame the Work Package Early Neutron Source (WPENS) of the EUROfusion Consortium, first preliminary deterministic analyses of selected Reference Accident Scenarios (RAS) involving the PLO have already been performed in the past [10,11,12,13].

References [10] and [11] analysed a Loss-Of-Flow Accident (LOFA) sequence caused by an EMP trip, using the MELCOR v1.8.6 code for fusion applications and the RELAP5-3D code, respectively. These works confirmed the importance of designing a reliable fast beam shut down system to avoid the occurrence of lithium boiling in the TVC, consequent potential damages and mobilization of tritium and activation products.

In [12], a Loss-Of-Coolant-Accident (LOCA) caused by a large break in the QT is studied with MELCOR fusion. Some metrics of interests were obtained, such as the spilled lithium inventory in the TC room (approximately 700 kg) and the temperature transient of the stainless-steel liner. The potential pressurization of the water cooling system serving the liner was also assessed.

In reference [13], the analysis of two Li fires, caused by two postulated initiator events involving Li spills and affecting two different rooms (Li sampling room and LLC room), has been carried out with a Fire Dynamics Simulator (FDS) code. The analysis was focused on obtaining the maximum temperature and pressure values reached inside each room.

In this work, two postulated events, i.e. breaks in the PLO, leading to a Li spill into the Test Cell room have been studied deterministically using the MELCOR fusion code. For the analysis of the first RAS, the original computational model of the IFMIF-DONES PLO, previously described in [10], has been updated according to the current LS design and conservative results (e.g. maximum temperature values reached by the TC room and liner) have been provided. For the study of the second RAS, a specific methodology has been proposed to better simulate the heat transfer between Li and the TC liner and to provide important metrics such as the overall Li solidification time.

First Studied RAS

The first scenario analysed with MELCOR involves a double-ended guillotine break at the outlet of the reducer nozzle of the TA, resulting in a consequent Li spill into the TC room. This initiating event is postulated as one of the selected IFMIF-DONES RAS involving the degradation of the PLO in correspondence of the Target Assembly [14]. For this scenario, a specific MELCOR model of the complete loop has been developed by updating the existing model used in [10] according to the current status of the LS design.

Model Description and Steady-State Initialization

The MELCOR model of the IFMIF-DONES lithium loop and ancillaries, shown in Fig. 3 and whose initial conditions are summarized in Table 1, is composed by 11 Control Volumes (CVs), 11 Flow Paths (FLs), and 15 Heat Structures (HSs). Specifically, 6 CVs are representative of the different parts of the loop: the straightener-reducer nozzle (CV100) and the TVC (CV200); the QT and the Li outlet pipe (CV300); the EM pump (CV350) and the heat exchanger (CV400), and, finally, the Li inlet pipe (CV450). In addition, the vacuum system serving the loop through the last section of the accelerator beam duct (CV150) is modelled as a boundary volume (CV050). Finally, the TC room, the Lithium Loop Cell (LLC) room and the outer environment are modelled as independent CVs (CV130, CV600 and CV700, respectively).

To simulate the Li jet flowing inside the TVC (the total Li mass inventory in the loop is, approximately, 3500 kg), two FLs (FL100 and FL200) have been implemented, while the 5 MW-thermal power deposited by the deuteron beam has been simulated as an energy source injected in the lithium residing inside the TVC. In addition, following the same approach described in reference [10], these FLs are forced to have the same mass flow rate by means of a specific control function.

The stainless-steel liner which covers the TC has been modelled by means of 3 HSs (with a thickness of 1 cm): HS13001 for the floor, HS13003 for the ceiling, and HS13005 for the walls. Finally, to simulate the HX, the temperature of the HS associated to CV400 (HS40001) has been adjusted so that the thermal power extracted is equal to the power injected in the TVC. This same solution has been adopted in reference [10] to overcome the current limitation of the MELCOR fusion code which does not allow the simultaneous use of multiple working fluids in the same code input.

Fig. 3
figure 3

MELCOR model of the DONES lithium loop and ancillaries

Full size image
Table 1 Initial conditions of the MELCOR model for the first RAS studied
Full size table

First, a steady state initialization of the code input has been run for the first 1000 s. The transient sequence starts at the time of 1000 s, when the break occurs. The break at the outlet of the straightener-reducer nozzle has been simulated by opening valves in FL700 and FL800 (in red in Fig. 3); at the same time, the valve in FL100 (which connects the straightener-reducer nozzle with the TVC) has been closed. Then, when Li starts to spill into the TC room, the signal for the beam shutdown is generated at time 1000.1 s. Table 2 describes the complete sequence for this simulation, including the main events of the start-up phase (the power injection occurs when the SU1, SU3, and SU4 conditions are reached), the normal operation, and the beam shut down, which occurs when one of the following conditions is met: the gas pressure in the TVC is higher than 1 kPa, the Li jet velocity is minor or equal than 5 m/s, the Li temperature inside the TVC is higher than 310 ºC and the Li level inside the QT is lower than 0.5 m (in the table only the condition that is fulfilled and effectively causes the beam shut down has been indicated).

Table 2 Complete sequence for the RAS analysed
Full size table

Results

According to the MELCOR code predictions, less than 4% of the total lithium mass inventory will be spilled inside the TC room (as shown in Fig. 4), resulting in the formation of a Li pool at the bottom of the TC floor with a final height of 16.4 mm. Due to the position of the break in the upper part of the PLO, only the column of Li above the break and the portion of Li that is dragged by the inertia of the fluid (the electromagnetic pump has no inertia) is spilled.

The pressure and temperature transient inside the TC are represented in Figs. 5 and 6, respectively. Regarding the temperature transient behaviour inside the TC, in the first moments of the accident sequence, the spilled Li is at maximum temperature of 521 K. As the heat transfer with the TC atmosphere and floor (modelled with HS13001 associated to CV130) occurs, the temperature gradually decreases until reaching the Li solidification point. At this point, the code automatically switches from non-equilibrium to equilibrium conditions for which the lithium pool and room atmosphere are forced to be thermally coupled. The pressure transient evolution inside CV100 (straightener-reducer nozzle) shows a fast depressurization of the loop when the pump is tripped, quickly matching the pressure of the TC atmosphere.

Finally, Fig. 7 depicts the temperature transient of the TC floor liner (HS13001) whose last node in direct contact with the spilled Li (node 9) reaches a peak temperature of approximately 430 K. Due to the imposed temperature applied to the left side of the HS (313.0 K), there is an important temperature difference between both sides of the HS of more than 200 K.

Fig. 4
figure 4

Lithium mass spilled inside the TC room (CV130)

Full size image
Fig. 5
figure 5

Pressure transient of the reducer nozzle (CV100), TC room (CV130) and TVC (CV200)

Full size image
Fig. 6
figure 6

Temperature transient of the spilled lithium (in black) and of the TC atmosphere (in red)

Full size image
Fig. 7
figure 7

Temperature transient of the TC floor liner

Full size image

Second Studied RAS

The second deterministic analysis has been performed on the RAS involving a break at the bottom of the Quench Tank (QT) with a Li spill into the Test Cell (TC) room.

In this second RAS, several code limitations were experienced concerning the Li behaviour and the different thermodynamic conditions applied by the code during the transient sequence. Due to the fact that no solid lithium is allowed as hydrodynamic material in the MELCOR CVH package, computation sometimes fails to converge when the extrapolation of the Li Equation of State (EoS) approaches the solid phase. Also, MELCOR fusion is programmed to switch from non-equilibrium to equilibrium thermodynamic conditions between vapour and liquid whenever the fluid’s temperature reaches the melting point.

In order to avoid this limitations, the analysis of this second scenario has been divided into two sequences. The first sequence consists of simulating the break in the QT and the Li spill assuming that the heat transfer between Li and the TC floor is dominated by natural convection (mostly laminar). The second sequence starts at the end of the Li spill when the molten metal pool is considered to be static at the bottom of the TC room and the predominant heat transfer mechanism is expected to be conduction. To address these two sequences, two different MELCOR models have been developed, one for each sequence.

Model Description

The model for the first sequence, represented in Fig. 8, consists of 2 CVs, modelling the QT (including the volume of the TVC, CV300), the TC room (CV130), and the TC floor (HS13001). This HS has an effective area of 5 m2 and a total thickness of 1.51 m divided into two layers: one layer of concrete with 1.5 m of thickness and a stainless steel liner with 0.01 m of thickness. An adiabatic boundary condition for the left side and a free convective boundary condition for the right side have been imposed. The break at the bottom of the QT is modelled by a FL (FL500) considering a rupture area of 10 cm2.

Fig. 8
figure 8

MELCOR model for the first sequence

Full size image

The model for the second sequence, shown in Fig. 9, is similar to the model for the first sequence. In this case, the CV modelling the QT has been omitted and a Li pool volume equivalent to the volume of the spilled Li from the first sequence has been added inside the TC room (CV130).

Fig. 9
figure 9

MELCOR model for the second sequence

Full size image

As for the first RAS analysed, a first steady-state calculation of 1000 s has been performed, so that, for the first sequence, the break occurs at time of 1000 s (by opening the valve in FL500, which models the break). Thus, the initial conditions for the model of the second sequence, regarding the values of the thickness and temperature of the Li pool, the temperature and pressure of the TC, and the HS nodes temperature, have been extracted from the previous sequence at time of 2500 s (1000 s of steady-state + 1500 s of transient), when the transient of the first sequence has terminated. The initial conditions for both sequences are summarized in Table 3.

Table 3 Initial conditions for both sequences
Full size table

Results

According to the computed results, MELCOR predicts that the QT takes 630 s to completely empty from the start of the accident sequence (at 1000 s), resulting in a liquid Li pool of 0.1185 m of height as it can be observed from Fig. 10.

Figure 11 shows the pressure transient evolution inside the QT and the TC. The TC room is initially pressurized with helium at 10 kPa, while the initial pressure inside the QT is controlled by the same vacuum conditions of the TVC (1 Pa in our computation). Due to this pressure difference, the QT volume experiences a gradual pressurization governed by the counter-current helium flow through the break between the two volumes until reaching the final equilibrium value when the QT is totally empty. At this moment, the pressure of both volumes decreases slowly due to the heat transfer occurring between the spilled liquid Li and the HS modelling the TC floor. Regarding the temperature transient behaviour inside the TC, Fig. 12 shows that the temperature of the TC atmosphere (in red) initially increases rapidly, reaches a peak value of 593 K and, then, decreases following the same thermal behaviour of the spilled liquid Li (in black).

Additionally, Figs. 12 and 13 show the temperature transient of the TC floor liner for sequences 1 and 2. Due to the different thermal conductivity of stainless-steel and concrete, the nodes of the floor liner rapidly match the temperature of the liquid lithium, while a slower temperature transient affects the concrete layer acting as a thermal insulator.

When the first sequence has been completed, the calculation of second sequence starts by applying initial conditions extracted from the previous phase at the time of 2500 s. Considering the temperature transient in the TC floor (represented in Fig. 13), MELCOR predicts a solidification time for the liquid Li spilled into the TC room of approximately 13 h.

Fig. 10
figure 10

Liquid Li pool level inside the TC room (CV130) for the first sequence

Full size image
Fig. 11
figure 11

Pressure transient in the TC room (black) and inside the QT (red) for the first sequence

Full size image
Fig. 12
figure 12

Temperature transient inside the TC room (CV130) for the first sequence

Full size image
Fig. 13
figure 13

Temperature transient of the TC floor for the second sequence

Full size image

Summary and Conclusions

A deterministic analysis with the MELCOR v1.8.6 for fusion code of two potential accident scenarios involving a Li spill inside the TC room has been performed, providing several metrics of interest, such as the maximum temperature and pressure loads in the TC room, the spilled Li mass, the maximum temperature reached by the stainless steel liner, and the Li solidification time.

The first studied accident scenario was a double-ended guillotine break of the inlet nozzle of the TA with a Li spill into the TC, for which a MELCOR model of a complete Li loop has been developed. The results predicted by MELCOR indicate that, after the loop rupture event, a small amount of the total available Li inventory of the PLO (less than 4%) is spilled in the room and rapidly cools down reaching the freezing point.

The second studied accident scenario was a break of 10 cm2 in the QT with a Li spill into the TC. To avoid the limitations of MELCOR, concerning the Li EoS extrapolation and the thermodynamic conditions at the Li melting point, the analysis has been divided into two sequences and two MELCOR models have been developed. The results obtained with MELCOR indicate that 630 s are required to empty the QT, spilling around 600 kg of Li into the TC (≈ 13% of the total Li inventory of the PLO). Finally, the time needed for the complete Li solidification inside the TC room was predicted as 13 h (approximately) for this case. This long time highlights the passive function provided by the room floor made of concrete which acts as a thermal insulator.

In conclusion, it has been possible to analyse the phenomenology and to obtain the main metrics involved in this type of accident scenarios. These scenarios need to be further analysed by refining and updating the current models as the facility progresses towards its final design (future changes will mainly affect the PLO design). The results of this work will also support the design of fire protection strategies to be applied to the IFMIF-DONES facility for preventing and/or mitigating accident scenarios which may involve Li spills and fires. For example, the maximum temperature and pressure loads calculated inside the TC room can be used as input for identifying the design requirements of the TC and the liner. Furthermore, other results provided by this analysis, such as the Li solidification time, can be critical for defining passive strategies to minimize the fire risk in case of potential Li spills as described in [15]).