This paper gives an overview of the study with emphasis on phenomena and their importance for processes resulting in activity releases.

The following phenomena were included as possible causes of activity releases from the containment:

• Hydrogen Deflagration/Detonation
• Direct Containment Heating
• In-vessel ans Ex-vessel Steam Explosions
• Global Vessel Failure
• Thermal Attack of Pentrations
• Basemat Failure due to Core Concrete Interaction

The Level 2 Study also includes a sensitivity analysis where the impact on the results from uncertainties in the phenomena are estimated. This is shortly described in this paper together with the effect of human errors.

Finally the results of the study are summarized and discussed.

The Phebus containment vessel is designed so that the temperatures of all the surfaces and the sump can be regulated: they can either be held constant or changed in a pre-determined manner. The experimenters then have to choose values of the surface temperatures that produce the desired humidity when steam and hydrogen are entering the vessel at a specified rate and afterwards when the aerosols are settling.

The first approach in determining the best surface temperature for the vessel was to run computer codes such as JERICHO and CONTAIN that are used for calculating the consequences of an accident in a full-sized reactor. The results of these calculations demonstrated a sensitivity to heat and mass transfer correlations. The codes have not been validated in geometries similar to the Phebus vessel so a priori there were few grounds for preferring one correlation to another.

In this paper we summarize blind pre-test calculations performed with the codes JERICHO, MELCOR, CONTAIN, CONTEMPT and CONT, show how the results compared with the thermalhydraulic experiments and go on to analyze the reasons for the discrepancies in the code results.

Armed with the knowledge of these tests another benchmark exercise was organized with boundary conditions corresponding to measured values from the experiments. Codes using the well-mixed hypothesis as well as multi-compartment and three dimensional codes were used. Comparing the results of the calculations with measurements contributed to further understanding of the thermalhydraulic behaviour of the Phebus-FP vessel and a better understanding of the uncertainties of the codes.

Criterion for spontaneous detonation onset possibility and its application to severe accident in a nuclear power plant is discussed. Theoretical and experimental results on spontaneous detonation onset conditions are summarised. Three series of large scale turbulent jet initiation experiments have been carried out in KOPER facility (50 m³ and 150 m³). Series of jet initiation experiments in initially confined H₂ - air mixtures have been carried out in KOPER facility (20 - 46 m³). Turbulent deflagration/DDT experiments were carried out in large scale confined volume of 480 m³ in RUT facility. Transition to detonation was observed at min. of 12.5% H₂. Results showed, that the characteristic volume size should be used for conservative estimates in accident analysis. Series of experiments on detonation transition from one mixture to another of lower sensitivity has been carried in DRIVER facility. The experiments were aimed on the estimation of the minimum size of a detonation kernel. The received results are in a good agreement with the 7 criterion.

3D computer codes 3ET and b02 have been developed for description of loads from detonations. Series of large scale H₂ detonation experiments have been carried out in RUT facility (16 - 25 %H₂, two initiator locations). Experimental results form a database on detonation loads on reactor-relevant scale and complex 3D geometry. B02 code was evaluated against experimental data. Good agreement of main loading parameters was observed for fully developed detonations. Effect of DDT location on loads has been studied experimentally (UTR facility) and numerically in 1D geometry. It was found, those peak overpressures depend strongly on DDT location, and can be significantly higher than that from detonation. Impulses depend mainly on the mixture volume. Data of large scale detonation, deflagration and DDT experiments (RUT facility) confirm these observations. Experimental data on turbulent flame propogation and on resulting loads (including DDT events) were received in large-scale RUT experiments.

Recent results of combined hydrogen injection/ignition experiments are presented. The experiments are aimed on the investigation of possible consequences of deliberate ignition at dynamic conditions. Experiments include the large-scale tests on the effects of igniter location, ignition time, injection rate (0.1 - 1 kg/s) and injection point on the combustion mode. The possibility of initiation of local detonations due to ignition at dynamic conditions was observed in the tests. The experiments showed a good local mixing and large-scale hydrogen concentration nonuniformities. Multiple explosions at continuous injection were observed. The possibility of local detonations resulting from deliberate ignition should be taken into account in accident analysis.

The propagation of a freely expanding flame is intrinsically unstable. Due to a feedback mechanism between the combustion induced flow and the combustion itself, a flame can accelerate very rapidly if obstructions are placed along its path. If appropriate conditions (in terms of the composition of the mixture, the flame speed and the configuration of the obstruction) are present, a DDT can occur. Presently, these conditions for H₂-air-steam mixtures have not been determined. This paper describes a systematic study of flame propagation in a duct filled with obstacles to identify the transition limit (in terms of mixture composition), the transition distances and the critical flame speed leading to a DDT. Experiments were performed in a 28 cm diameter, 6-m-long combustion duct. Baffle type obstacles, having a blockage ratio (blocked area to cross-sectional area ratio) of 0.4 were mounted along the duct to induce turbulence in the unburned gas. Piezoelectric pressure transducers were mounted along the wall of the duct to monitor the location of a DDT as well as flame speed just prior to the transition process. Results showed that the range of mixture (in terms of the H₂ concentration) for which DDT has been observed reduces as the steam concentration increases. DDT was not observed in any mixture containing more than 30% of steam. Results also showed that DDT did not occur if a flame had not accelerated to a speed corresponding to a flame Mach number greater than 1.5. The transition limits and the critical flame speed are necessary conditions for DDT; both of these criteria have to be satisfied before a DDT can occur.

The essential feature of the AECL recombiner is a novel catalyst material which is unmatched in wetproofing and thermodynamic range of operation. The catalyst is the product of the 20 years of engineering of high-quality catalysts to separate hydrogen isotopes in heavy water manufacture, a continuing key technology area in AECL. The catalyst is a nuclear product manufactured exclusively by AECL. It is successfully used to control hydrogen in other nuclear applications such as in the transport and storage of wet radioactive materials, and in fuel reprocessing operations. The active catalyst material is platinum, matrixed in an inorganic wetproofing support and bonded to a stainless steel substrate by a proprietary process. The formulation has extraordinary stability and robustness. Problems such as spalling and evaporation of wetproofing at high temperatures are eliminated. For this application, the catalyst is manufactured in thin, tough sheets with low mass for fast heat dissipation and low bulk for compact size per unit capacity.

Function and performance of the AECL recombiner were demonstrated in the 6.6 m³ and 10.7 m³ Containment Test Facility (CTF) vessels at AECL Whiteshell Laboratories (WL) using a 1/10-scale test model and a full-scale prototype recombiner. The AECL recombiner removes hydrogen (and carbon monoxide) at a high capacity and is resistant to foreseeable poisons and fouling agents in containment, during normal operation, or in an accident.

The AECL recombiner is self-starting in the range of temperatures 20 to 150^oC at a low H₂ concentrations (<1.0% V) and starts at low temperatures in a condensing atmosphere. Humid performance at low temperatures is considered a vital test of wetproofing effectiveness and is critically important in containments where pressure suppression systems are employed.

The nominal capacity for hydrogen recombination is 5.0 0.2 kg/h per m² of inlet area to the recombiner in 5.0% H₂ in air at 25^oC. The capacity is increased by approximately 30% per metre of chimney length. The capacity is insensitive to the presence of the diluents such as steam, CO₂, or N₂ as long as the minimum stoichiometric amount of oxygen is available to recombiner the hydrogen. The capacity increases directly in proportion to the initial pressure in the range 1-2 atm. The capacity increases in proportion to the concentration of the limiting reactant. Finally, the capacity for hydrogen removal increases in proportion to the inlet area to the recombiner housing. This simple scaling parameter was verified in identical tests with the 1/10-scale test model and full-scale prototype recombiners.

Specific experiments proposed in the initial program are pertinent to (a) the interaction of corium melt and the lower head of the pressure vessel, (b) the interaction of corium melt with in-vessel and ex-vessel water pools. We expect that experiments on in-vessel ex-vessel melt and debris bed coolability will also become a part of this program of experiments on melt-structure water interactions.

The first set of experiments performed investigate the ablation process in a LWR vessel, as the melt discharges from the vessel to the containment through the failure location, be it a penetration, or the initial opening during through the creep-rupture process. The simulant material chosen is a mixture of Pbo and B₂O₃ which melts at ~1000 K and has very low viscosity at 1000 K and above. The vessel material chosen is lead, which melts at 600 K. The objective of these experiments is to determine if a crust formed retards the heat transfer to the vessel wall or is swept out by the flowing melt and is ineffective. The extent of the final size of the hole is affected significantly by the presence or absence of the crust.

Scaling of the hole ablation process has been performed. It appears that the experiments performed with a ? 10 to 100 l of melt will cover the values of the parameters for the prototypic accident conditions. A one dimensional code for melt flow and vessel interaction, called HAMISA (hole ablation in severe accidents) has been completed, and a two dimensional code is being prepared. The process is basically two dimensional, as has been demonstrated in the experiment with a thick plate, having an initial hole of 10 mm.

As far as the containment is concerned, mitigation features allowing to limit pressure and temperature inside the building are to be assessed. There is also a need to accurately to estimate local hydrogen concentrations in order to evaluate the hydrogen risk.

For this purpose, codes have been developed in the past twenty years around the world. they generally include models for all the physical phenomena involved during severe accidents. Most of them are based on a multi-compartment approach for the spatial description. This approach have well known limitations when local concentrations must be evaluated in large volumes.

On the other hand, general purpose multi-dimensional thermal hydraulics computer codes are able to predict complex situations such as stratification occurrence, but they often lack some models which are necessary to be included in the analysis of the containment under a severe accident. The ability to treat long transients is also a concern with this type of codes.

For these reasons, CEA/DMT has undertaken the development of the GEYSER/TONUS code. This code will allow the coupling of parts of the containment described in a lumped parameter manner, together with meshed parts. Physical model of classical lumped parameter codes, such as condensation, will be adapted for the spatially described zones. The objective is to be able to treat complete scenarios and optimized numerical methods are developed for the transient problem. The code is built in the CASTEM 2000/TRIO EF system which allows, thanks to its modular conception, to construct sophisticated applications.

In this paper, the GEYSER/TONUS code is described, and some applications are shown.

The design of an ad-hoc filter requires consideration of :

• availability of materials;
• ease of construction and operation;
• stability of materials under heat and irradiation;
• filter efficiency (aerosols and vapours);
• filtration capacity;
• decay heat removal;

The focus here will be on generic aspects of heat and mass transfer in ad-hoc filter design. Key filter parameters will be identified and discussed in terms of projected loadings on the filter. The advantages and disadvantages of alternative materials and designs will be considered.

Furthermore, iodine has volatile chemical forms, which increases the probability of its release to the environment. However, iodine chemistry is quite complex, and the behaviour of iodine under accident conditions can be influenced by numerous variables. For example, the extent to which CsI, the most likely form of iodine to enter containment, would undergo reaction to form volatile I₂ or organic iodides, is dependent on the dose rate, solution pH and impurities present within containment. Surfaces also play an important role in influencing iodine volatility, both as a sink and as a source of impurities that can increase iodine volatility.

This paper highlights recent findings in iodine research and summarizes the current state of understanding of iodine chemistry relevant to reactor safety. This paper also puts forward recommendations for minimizing iodine release from containment.