Methods and Data for HTGR Fuel Performance and Radionuclide Release Modeling During Normal Operation and Accidents for Safety Analysis

FORSCHUNGSZENTRUM JÜLICH GmbH KFAJ

Institut für Sicherheitsforschung und Reaktortechnik

Methods and Data for HTGR Fuel Performance and Radionuclide Release Modeling during Normal operation and Accidents for Safety Analysis

K. Verfondern R. C. Martin R. Moormann Berichte ties Forschungszentrums Jülich ; 2721 ISSN 0366-0885 Institut für Sicherheitsforschung und Reaktortechnik Jül-2721

Zu beziehen durch : Forschungszentrum Mich GmbH - Zentralblbllothek Postfach 1513 - D-5170 Mich - Bundesrepublik Deutschland Telefon: 02461f61-6102 - Telefax : 02461161-6103 - Telex; 833556-70 kfa d

Methods and Data for HTGR Fuel Performance and Radionuclide Release Modeling during Normal Operation and Accidents for Safety Analysis

K. Verfondern') R. C. Martin 2) R. Moormann')

Research Center Jülich 2) Oak Ridge National Laboratory

Research Center Jülich GmbH - 1SR Jül - 2721 January 1993

METHODS AND DATA FOR HTGR FUEL PERFORMANCE AND RADIONUCLIDE RELEASE MODELING DURING NORMAL OPERATION AND ACCIDENTS FOR SAFETY ANALYSES

K. Verfondern (Research Center Jülich) R. C. Martin (Oak Ridge National Laboratory) R. A7oortnann (Research Center Jülich)

ABSTRACT

The previous status report released in 1987 on reference data and calculation models for fission product transport in High-Temperature, Gas-Cooled Reactor (HTGR) safety analyses has been updated to reflect the current state of knowledge in the German HTGR program. The content of the status report has been expanded to include information from other national programs in HTGRs to provide comparative information on methods of analysis and the underlying database for fuel performance and fission product transport. The release and transport of fission products during normal operating conditions and during the accident scenarios of core heatup, water and air ingress, and depressurization are discussed .

Forschungszentrum Jülich GmbH - ISR Jül - 2721 Januar 1993

RECHENMETHODEN UND DATEN ZUM HTR-BRENNSTOFFVERHALTEN UND ZUR. SPALTPRODUKTFREISETZUNG IM NORMALBETRIEB UND STÖRFALL IM RAHMEN VON SICHERHEITSANALYSEN

voll

K. Verfondern (Forschungszentrum Jülich) R. C. Martin (Oak Ridge National Laboratory) R. Moormann (Forschungszentrum Jülich)

KURZFASSUNG

Der im Jahre 19137 erschienene Statusbericht mit der Beschreibung einer Referenz-Datenbasis und Rechenmodellen ist auf den neuesten Stand gebracht worden und beschreibt den gegenwärtigen state-of-the-art im deutschen HTR-Programm . Der Inhalt des Statusberichts ist erweitert worden um Informationen aus den HTR-Programmen anderer Nationen, um einen Vergleich der Analysemethoden sowie der zugrunde liegenden Datenbasis zur Beschreibung des Brennstoffverhaltens und des Spaltprodukttransports zu ermöglichen . Der Bericht umfaßt die Bereiche Freisetzung und Transport von Spaltprodukten während des Normalbetriebs sowie im Verlaufe von Unfallszenarien von Kcrnauflieizung, Wasser- und Lufteinbruch und Druckentlastung.

Table of Contents

1 .0 INTRODUCTION ...... I

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION ...... 3 2.1 FUEL DESIGN ...... 3 2.1 .1 Coated Particle ...... 3 2.1 .2 Fuel Element ...... 10 2.1 .3 HTGR Fuel Quality ...... 11 2.1 .4 Fission Product Inventories ...... 13 2 .2 COATED PARTICLE IRRADIATION PERFORMANCE ...... 13 2.2.1 Kernel Migration (Amoeba Effect) ...... , ...... , . 17 2.2.2 Fission Product Interaction With Silicon Carbide ...... 18 2 .2.3 Pressure Vessel Failure ...... 19 2 .3 RADIONUCLIDE RELEASE DURING NORMAL OPERATING CONDI- TIONS ...... 19 2 .3 .1 Metallic Fission Product Release ...... 22 2.3.2 Uptake of Fission Product Metals by SiC , ...... 24 2 .3 .3 Fission Gas Release ...... 25 2.3 .4 Fission Product Activity Distribution in the HTGR Primary Circuit at Acci- dent Initiation ...... 32

3.0 HTGR FUEL BEHAVIOR DURING CORE HEATUP ACCIDENTS ...... 41 3 .1 THERMODYNAMICAL BOUNDARY CONDITIONS ...... 41 3.2 COATED PARTICLE PERFORMANCE UNDER ACCIDENT CONDI- TIONS ...... 42 3 .3 RADIONUCLIDE RELEASE FROM COATED PARTICLES UNDER AC- CIDENT TEMPERATURE CONDITIONS ...... 47 3.3 .1 Metallic Fission Product Release ...... 47 3.3.2 Fission Gas and Iodine Release ...... 51 3.3.3 Particle Failure Model Discussion ...... 53 3.4 FISSION PRODUCT TRANSPORT WITHIN THE CORE CAVERN . . . . . 54

4.0 HTGR FUEL BEHAVIOR DURING WATER AND AIR INGRESS kCCI- DENTS ...... 61 4.1 Fission Product Release from Defective Coated Particles ...... 61 4.2 Fission Product Release from Graphite ...... 63 4.3 Massive Long Term Air Ingress ...... 66 4.4 Release of Fission Products Plated-Out on Metal Surfaces ...... 68 4.4.1 Mobilization by Liquid Water ...... 68

Table of Contents iii

4.4.2 Mobilization by Steam or Air Attack ...... 69 4.4.3 Exemplary Source Term Contributions Caused by Mobilization of Plateout Activity ...... 73 4.5 Fast Reactivity Transients in Combination with Water Ingress ...... 74

5.0 HTGR FUEL BEHAVIOR DURING DEPRESSURIZATION ACCIDENTS . 75 5.1 Desozption of Plateout Activity due to Pressure Drop ...... 75 5.2 Liftoff of Dust-Borne Activity ...... 75 5.3 Source Terms in Depressurization Events ...... 80

6.0 ACKNOWLEDGEMENT ...... 83

7.0 REFERENCES ...... 85

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL ...... 105 A. I FUEL KERNEL ...... 107 A.2 PYROCARBO'N ...... l08 A.3 SILICON CARBIDE ...... 110 A.4 GRAPHITE ...... 113 A.4.1 Matrix Graphite ...... 113 A.4.2 Structural Graphite ...... 116 A.4.3 Concentration Dependence of the Diffusion Coefficient ...... 117 A.5 ZIRCONIUM CARBIDE ...... 120

Appendix B. INPUT DATA FOR PARTICLE FAILURE MODELS UNDER AC- CIDENT CONDITIONS ...... 141

Appendix C. SORPTION ISOTHERMS OF FISSION PRODUCTS OVER GRAPHITIC SURFACES ...... 149

LIST OF ABBREVIATIONS

AGR Advanced Gas-Cooled Reactor AVR Arbeitsgemeinschaft Versuchs-Reaktor BISO Buffer Isotropic Coating (Buffer and Pyrocarbon Layers) BOL Beginning-of-Life CAGR Commercial Advanced Gas-Cooled Reactor CEGB Central Electricity Generating Board EFPD Equivalent Full Power Day EOL End-of-Life FIMA Fissions per Initial Metal Atoms FRG Federal Republic of Germany FZ Forschungszentrum (Rossendorf) GA General Atomics HFR High Flux Reactor (Fetten) HOBEG Hochtemperaturreaktor- Brennelement GmbH HRB Hochtemperatur-Reaktorbau GmbH HTGR High-Temperature Gas-Cooled Reactor HTI High-Temperature Isotropic HTTR High-Temperature Engineering Test Reactor IAEA International Atomic Energy Agency INET Institute for Nuclear Energy Technology JAERI Japanese Atomic Energy Research Institute KFA Forschungszentrum (Jülich) KfiFA Kühlfinger-Versuchsapparatur LEU Low-Enriched Uranium LTI Low-Temperature Isotropic MHTGR Modular High-Temperature Gas-Cooled Reactor MIT Massachusetts Institute of Technology MTR Material Test Reactor SEM Scanning Electron Microscopy THTR Thorium-Hochtemperaturreaktor TRISO Tristructural Isotropic Coating (Buffer, SiC, PyC) VGM Modular HTGR Design of the Russian Federation VHTR Very High Temperature Reactor

LIST OF ABBREVIATIONS v vi 1.0 INTRODUCTION

Much experience has been gained since the 1960s in the operation of high- temperature, gas-cooled reactors (HTGRs), beginning with the experimental Peach Bottom reactor in the United States (US), continuing with the Dragon reactor project in the United Kingdom (UK) and the AVR project in the Federal Republic of Germany (FRG), and resulting in the construction and operation of the prototype reactors Fort St. Vrain in the US and the Thorium High- Temperature Reactor (THTR-300) in Germany for the production of electricity. Unfortunately, for a combination of technical, political, and economic reasons, these reactors are no longer in operation and no more are planned in the near future in Germany.

The experience with these reactors has consistently demonstrated the safety margins inherent in HTGR design, even under adverse circumstances and accident scenarios. The safety features and the economic viability of compact reactor designs continue to attract international interest in the HTGR concept. Although the German HTGR program has a long and productive history, the HTGR as a national priority has been deemphasized for the future. The US maintain an active research program in the modular-HTGR (MHTGR) design, but a long- term commitment to the construction of new HTGRs has not been made at the present time.

In contrast, other nations have expressed a serious near-term commitment to construction of small demonstration HTGRs. Construction of the 30 MW(th) High-Temperature Engineering Test Reactor (HTTR) is currently underway in Japan . The People's Republic of China is moving ahead with plans for a 10 MW(th) Test Module HTGR to be constructed later this decade. HTGR-related experimentation and research continues in other countries, most notably the Russian Federation, the UK, and France.

The changing national priorities in HTGR development suggest that international cooperation in documenting the existing design database and analysis of existing models and codes used in safety studies could enhance the prospects for future licensing of HTGR reactors around the world. Each national HTGR program has specific strengths and emphases, and comparison of the current international state of knowledge can be a cost-effective approach to defining pri-

1.0 INTRODUCTION ority data and analysis needs as well as providing a cooperative forum for validation of predictive safety codes using data from present or future experiments.

In 1987, a status report on reference data and models for fission product transport to be used in German HTGR safety analyses was issued (Ref. 1). This status report summarized the German experience in the development and quality assurance of fuel, to be used as the basis for validating calculational models which can reproduce experimental results and predict the fission product release behavior under normal operating and accident conditions.

The status report presented here updates Ref. 1 to include new data and models which are currently used in HTGR studies. To consider the international experience in HTGRs as discussed above, this report has also been expanded to include information on the HTGR data and models in other national programs. The German experience is presented in the most detail, followed by US and Japanese methodology and that of other countries as available.

This report does not suggest which models or supporting data should be recommended, but rather attempts to describe the models as they exist and to comment where appropriate on areas in which limited knowledge may require conservative modeling approaches. By summarizing the current state of HTGR safety analysis methodology, this report can hopefully provide guidance in future improve- ments in this methodology and define areas in which future R&D work could result in improved models.

Discussion of fission product transport is limited to the radiologically important species. The metals cesium and strontium are emphasized, with some discussion of silver. The fission gases krypton, xenon and iodine are emphasized. Informa- tion is presented for both for normal operating conditions and accident conditions, with the accident condition scenarios subdivided into core heatup, water and air ingress, and depressurization. Both expected values and design values for fission product release are presented for different scenarios. Discussion of accident consequences and the dispersion of fission product outside the primary circuit is beyond the scope of this report.

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION

2.1 FUEL DESIGN

2.1 .1 Coated Particle

The fundamental characteristics of fuel for HTGRs represented by ceramic coated particles have been investigated for thirty years. Several countries have initiated a fuel development and qualification program with the coated particle as the basic unit.

In the Federal Republic of Gennany (FRG), the production process for spherical fuel elements with high-enriched (Th,U)02 particles with a BISO coating as fuel for the AVR and THTR reactors was fully established and licensed in the 1970s . In the early 1980s, TRISO coated particles with lowenriched uranium (LEU) were chosen as reference fuel for new HTGR designs. These particles consist of a 10.6 % enriched U02 kernel with a diameter of 500 ym surrounded by subsequent layers of buffer (thickness: 95 um), inner pyrocarbon (iPyC, 40 um), silicon carbide (SiC, 35 ,um), and finally outer pyrocarbon (oPyC, 40 ,um) as shown in Fig. 2-1 . The qualification program for the LEU reference fuel has been ad- dressed to irradiation and accident conditions confirming the "1600 ° C concept" (meaning that this temperature limit will not be exceeded in any conceivable accident scenario) as a new passive safety concept for future German modular HTGRs (Ref. 2) . In the final step of the qualification program, the so-called proof tests, full size fuel elements are being tested in the Material Test Reactor (MTR) HFR Petten under simulated HTGR operating conditions. Testing limits for modern German HTGR fuel have been chosen to envelope the required normal operating conditions. Values reached so far in various irradiation experiments are a burnup of 15 % FIMA (design limit: 8-10 %), a fast neutron fluence 25 .2 of 8* 10 m , E > 0.1 MeV (design limit: 2-3* 1025), and an irradiation temperature of 1250 °C (Ref. 10) . In 1989, the German manufacturing company HOBEG decided to abandon all production and development activities as a con- sequence of bribary scandals in the German fuel cycle activities(Ref. 2) .

The US reference fuel design is based on a two-particle concept of fissile and fertile particles (Fig. 2-2). The fissile particle consists of a fuel kernel (diameter

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 3 Fig. 2-1 : FRG Reference Design for Fuel Particles and Fuel Element

350 hem) containing 19.6 % enriched uranium oxycarbide (UCO) which is surrounded by a 100,um buffer layer, 50 ,um iPyC layer, 35 ,am SiC layer, and 40 ,um oPyC layer. The fertile TRISO particle has a 500 ym diameter Th02 kernel surrounded by 65 um buffer, 50 um iPyC, 35 ,um SiC, and 40 um oPyC layers. The conditions required for US fuel are more severe than German fuel . Design limits here are a burnup of 26 % FIMA for the fissile particle and a fast fluence of 4.5*102$ m-2, E > 0.18 MeV' (Ref. 3) . The latest reference US fuel was recently tested in the HFIR reactor at the Oak Ridge National Laboratory (ORNI.) in the irradiation experiment HRB-21 . In a recent plant cost reduction

Conversion factor for the unit of the fast neutron fluence from E > 0.18 MeV to E > 0.1 MeV is 1 .1.

US D

Fissile .-a-360 mm --

U-C-0-Kernel

E Fertile m

Th02-Kernel

Fuel Particles Fuel Rod Fuel Element

Fig. 2-2: US Reference Design for Fuel Particles and Fuel Element

study for the MHTGR, a major change was suggested to be the replacement of thorium with natural uranium which is expected to reduce the maximum fuel temperature under core heatup accident conditions (Ref. 4).

The reference coated particle design in Japan for the first HTTR core has been completed using low enriched (3 .3 - 9.9 %, average 6 %) U02 fuel (Fig. 2-3). The TRISO particle has a kernel diameter of 600 um and layer thicknesses of 60 /cm, for buffer, 30 pm for iPyC, 25 pm for SiC, and 45 ,um for oPyC. The main differences compared to the FRG and the US design are a larger particle kernel and a thinner silicon carbide interlayer. The burnup design limit for the HTTR reference fuel is as low as 3.6 % FIMA. For the HTTR second core, efforts are being put into the development of "advanced fuel" covering the improvement of the fabrication process, the optimization of the particle design, for instance a thicker SiC layer to allow for a higher burnup, and the development of a ZrC coating (Ref. 5) .

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 5

Fuel Compact Fuel Rod Fuel Block

Fig. 2-1 : Japanese Reference Design for Fuel Particles and Fuel Element

The reference particle design in the Russian Federation for its version of a 200-250 MW(th) small modular HTGR, called VGM, consists of a TRISO coated particle with a 500 ym LEU U42 particle kernel with 8 % (values range between 6.5 and 10.0 %) enriched uranium, a 90 - 100 um buffer layer, a 70 - 80 um iPyC layer, a 60 urn SiC layer and a 60 jim oPyC layer. The burnup design limit is 10 FIMA. There is a broad experience in reactor experiments with Soviet fuel even under sharp pulse irradiation conditions. Fuel has been tested so far in the temperature range 800-1600 °C up to 625 equivalent full power days (efpd) with maximum burnups of 10 - 15 % FIMA and fast fluences up to 2.3* 1025 m-2, E>0.18 MeV. (Ref: 6) .

Also in the Institute for Nuclear Energy Technology (INET, Tsinghua Universi- ty, Beijing) R&D program in China for an HTR fuel element to be inserted in the 10 MW(th) HTR Test Module, TRISO coated fuel particles have been manufactured with the following nominal geometrical data with only a small difference compared to the design data given in table 1: 501 pm kernel diameter, and layer

thicknesses of 92 um for buffer, 39 ym for iPyC, 36 um for SiC and 34 ,um for oPyC (Ref, 78). The fuel loading is 5 g uranium per fuel ball with an enrichment of 18 %. Preliminary operational design data are a maximum burnup of about 14 % FIMA (average: 9 %), a fast neutron fluence up to 1 .8* 1025 m-2 with a surface fuel temperature of 900-1000 °C. At present, particles produced on a laboratory scale were inserted in cold experiments with burnups < 1 % FIMA (Ref. 7).

As can be seen above, BTSO coated fuel particles are not a part of any country's HTGR fuel reference design. BISO fuel was used in the German real-time operating HTGRs AVR and THTR-300, both of which have terminated their operation in the meantime . It is planned to eventually investigate irradiated BISO fuel in heating tests at KFA Rilich . In Japan, BISO coated fuel particles are also used in experimental investigations.

Characteristic data on each country's coated particle reference design are summarized in table I .

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 7 .

" G "

A w

A CC ,.

:tf a F a cz ü ~ ä boo S ZZ t)

V ö " -0 0 'v Q I

C) co bmw ,. ., b W ;j v G. o cC C a Cs O :; o &- w U 4 w = ca. U m ..E iin 0 a, 9 F C7 2

. o0 .~ U ö

Q L,r) 00 Q Vi n u

.fl

U ° C7 ° 0 en En ~es Sx .-.. OQ N

A c~s C q4

_ ccs 09 u CV F~ ~ C

Ö kd') := CV v]

F a v C ~ ü > E ~ \ u GL ä 4- 7 F Q ~ m zw

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 9

2.1 .2 Fuel Element

Two main fuel element concepts are presently in use, the spherical fuel element and the block-type fuel element.

The FRG fuel element design (Fig. 2-1) is a graphite sphere with a diameter of 60 mm. The reference fuel sphere contains approximately 11,000 TRISO coated fissile particles. The particles are overcoated with an A3 matrix graphite layer with a thickness of 200 ym to prevent direct contact of the particles and then embedded in the same A3 matrix graphite material. The outermost 5 mm of the fuel sphere is a shell of matrix graphite only without any particles. The heavy metal loading of the reference fuel element is 0.5 g fissile (U-235) and 6.5 g fertile (U-238) material. The active core of the German 200 MW(th) HTR-Module consists of about 360,000 spherical fuel elements.

The US fuel element design (Fig . 2-2) is a hexagonal graphite block (793 mm in length and 360 mm wide across the flat surface) of the type 1-1-451 containing 102 coolant channels (diameter 15.9 mm) and 210 fuel holes which are filled with fuel compacts (formerly referred to as rods) and sealed. The fuel compacts (diameter 12 mm) are a mixture of TRISO coated fissile and fertile particles and graphite shim particles bonded by a carbonaceous matrix. An additional step of over- coating of the particles with an outer buffer layer of 40 - 60 pm thickness has been recently added to the fuel element manufacturing procedure, and was found to significantly reduce the manufacturing defect fraction (HRB-21). A small radial gap of 0.13 mm exists between the fuel compact and the fuel hole. The active core of the 350 MW(th) MHTGR consists of 660 graphite fuel elements.

Japan's fuel design (Fig. 2-3) consists of block-type fuel, similar to the US design, Each hexagonal graphite block (580 mm in length and 360 mm wide across the flat surface) has 31 or 33 fuel holes, each containing an annular fuel rod (inner diameter 26.3 mm, outer diameter 34 mm) which consists of 14 fuel compacts in a graphite sleeve (pin-in-block type fuel) . Fuel element block and fuel rod sleeve consist of the graphite IG-110 . A fuel compact made of graphite matrix powder with the shape of an annular cylinder (39 mm tall, inner diameter 10 mm, outer diameter 26 mm) contains 13,500 TRISO coated fissile particles which are overcoated similar to the German design . The coolant flows through a gap between fuel rod and fuel hole . The HTTR active core is composed of about 70,000 fuel compacts.

The Russian Federation has also chosen spherical fuel elements with 60 mm diameter including a 5 mm fuel-free graphite shell as reference design for the 200 MW(th) HTGR of modular type, called VGM. The main components of the matrix material are the artificial graphites on the basis of calcinated (30PG) or uncalcinated (MPG-6) graphite. Its active core is planned to be composed of about 350,000 fuel elements.

2.1 .3 HTGR Fuel Quality

The as-manufactured quality of HTGR fuel is expressed in terms of the fraction of accessible or free uranium; this includes the uranium present as heavy metal contamination outside particles with intact SiC layers and the uranium in particles whose kernels are exposed, i.e. particles in which all coating layers are frac- tured or permeable.

The large-scale production experience in the FRG has led to a high quality , spherical fuel element with a mean defect fraction of 3* 10-5 (102 detected defects out of 3.3 millions coated particles) (Ref. 12) . This corresponds - in the average - to only one defective particle every 2 - 3 fuel elements. Almost all free uranium found by the burn-leach technique can be attributed to defective particles, i.e. particles with exposed kernels (Ref. 13) . Only traces of natural uranium are found as heavy metal contamination of the matrix graphite, close to the detection limit of around 10-s. Thus the free uranium is less than the upper design limit of 6*10-5 guaranteed by the former German manufacturing company HOBEG. A fissile uranium fraction of 1 *10-7 has been assumed in KFA safety analyses calculations for the HTR-Module fuel elements. A design limit for the heavy metal contamination was set at 50 ,ug per ball (Ref. 16) corresponding to a fraction of about 7*10-7.2,

For Japan's fuel, a mass scale production on the order of 200 kg uranium per year started in 1983, and is planned to be doubled in 1992 for HTTR fuel pro-

2 In the previous status report (Ref. 1), the value assumed for the total free uranium fraction was the same both for expected and for design conditions . The single contributions, however, from heavy metal contamination of the matrix graphite and from defective coated particles were each defined in Ref. I to be half of the sum.

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION i 1 duction. The fraction of particles with a defective SiC layer in a fuel element expressing the quality of the fabricated fuel has been measured to be on the order of 5*10-4 (Ref. 72). Modified fabrication methods could even reduce this fraction to 3.9* 10-6 (Ref. I I). The heavy metal contamination in the fuel compacts was found to be around 3*10-5. It could also be reduced by almost one order of magnitude when using an improved fabrication process. Compared to these val- -3 ues, the regulatory requirement for the total amount of free uranium of < 2* 10 includes a large safety margin . This upper limit has been defined with respect to the maximum allowable radiation exposure of the public. The expected free uranium fraction, however, is 5.5* 10`4 (see table I1) .

In contrast, US fuel quality considers manufacturing defects of individual coating layers. A different modeling approach is assumed for particles with excessive heavy metal dispersion, for those with missing iPyC layers, for those with initially defective SiC layers, for those with missing buffer layers, and for those with missing or initially failed oPyC layers, with the sum of each fraction giving the total fraction of particles with manufacturing defects. The determination of the fraction of particles with specific defects is strongly dependent on the experimental techniques to measure them. The burn-leach technique is used for measurement of both the heavy metal contamination and the fraction of particles with failed SiC coating, in total the fraction of "free" uranium. A wet technique is used to measure the uranium content outside particles with an intact oPyC layer which is treated as heavy metal contamination. US specification limits are 1 * 10-5 for the heavy metal contamination in the fuel compact matrix material and 5* 10 5 for the fraction of particles with a defective SiC layer (Ref. 8) . The sum (b*10-5} representing the design limit for the fraction of free uranium is the same as in the German design .

Data characterizing the required as-manufactured quality of the Russian Feder- 10,6 ation's modular HTGR fuel are for the VGM: a U-235 contamination of 4* and a fraction of defective coated particles of S 10-4 (Ref. 9) . Besides their "tra- ditional" technology for fuel manufacture, the USSR have developed a second one using a pyrocarbon binder which is said to provide better mechanical properties but a higher contamination level in the graphite. The Soviet fuel production is on a laboratory scale so far.

The manufacture procedure for spherical fuel elements in China is also adopted from the German procedure but uses Chinese natural graphite. The fraction of

free uranium is surprisingly low at 4* 10'6 (Ref. 78). Particle and fuel production is still on a laboratory scale.

For former reactor designs which used BISO coated fuel particles (THTR-300, AVR), the fraction of free uranium in the active core differ from those of TRISO particles. In contrast to TRISO particles, the manufacturing process of B1SO fuel introduces a high heavy metal contamination fraction into the matrix graphite (3*10-) and into the HTI layer of the BISO coating (9* 10"~ due to high deposition temperatures > 2000 °C (Ref. 49). On the other hand, there are almost no particle defects due to manufacture (< 10"4), significantly lower than the design limit specification of 6* 10-4.

2.1 .4 Fission Product Inventories

The code OBIGEN (Ref. 50) is generally accepted as the standard for calculating the amount of fission product species and their distribution across the fuel zone. Test calculations have shown that the fission product inventories are being calculated with sufficient accuracy for HTGR safety analyses.

2.2 COATED PARTICLE IRRADIATION PERFORMANCE

Testing of HTGR fuel during irradiation has been conducted both in real-time operating HTGRs and in various MTRs . Testing limits have been chosen to provide a broad envelope around the required normal operating conditions of HTGR designs.

Within the FRG U0 2 LEU TRISO irradiation program, six tests have been made comprising a total of 260,420 coated particles. The use of different techniques for detection of defective/failed particles (cold gas test, cesium profile, chlorination) leads to slight differences in the results with the cesium profile method tending toward higher numbers. The comparison of the measured Kr-88 RIB-values before (BOL) and after (EOL) irradiation, however, has shown that no in-pile par-

2.0 HTGR FUEL BEHAVIOR DURING NORIVIAL OPERATION 13

title failure occurred. All detected defects could be concluded to originate from the manufacturing process (Ref. 10). Using statistical analysis, an upper limit for the irradiation induced failure fraction is represented by 2* 10'5 at the 95 % confidence level. No additional particle failure is expected for burnups up to approximately 9 % PIMA to assure the design limit chosen at 2* 10"4. In the status report (Ref 1), the design value was the same, whereas the expected value was previously chosen to be half of the design value.

During normal operation of the US-M HTGR, an additional fuel particle fraction of 6.1 * 10-5 is assumed to have a failed SiC layer. The fraction of 5* 10-5 particles with initially missing buffer layers has completely turned into exposed kernels (Ref. 8, see also section 2.2.3) . The US program also uses a fuel performance model for irradiation induced failure of the oPyC layer. The fraction of particles the oPyC layer of which has failed is as large as 3* 10-2 (Ref. 8). A qualification program similar to the German one has recently started, in order to demonstrate the quality goals for modern US fuel.

The performance of Japan's fuel under normal operating conditions has been investigated in many irradiation experiments to verify the safety design requirements of HTTR fuel; i.e., no additional systematic failure is expected to occur during normal operation with a maximum design fuel temperature of 1495 °C (nominal value ~_- 1300 °C). An upper regulatory defect limit for as- manufactured coated particles was set at 2* 10-3 at the 95 % confidence level (Ref. 11). Due to this high design value, no higher level of failure during normal (and abnormal) transients is expected to occur. However, for purposes of conservativism and considering a procedure similar to LWR licensing, a total defect fraction of 1 % is used in HTTR safety evaluation studies. A value of only 10-4 5* was experimentally found in compacts irradiated to 1 .5 % FIMA and reaching a temperature of 1750 °C. The maximum failure fraction increased by about two orders of magnitude compared to the initial value in irradiation experiments where a burnup of 4 % FIMA at 1720 °C was reached (Ref. 178).

Two major failure mechanisms during normal operation have been studied in detail in the Japanese program : kernel migration inside the coated particle (see section 2.2.1) and interaction of the fission product species palladium with silicon carbide (see section 2.2.2) . Neither one has been experimentally found to cause any further damage to the fuel under HTTR design operating conditions . R/B

values of Kr-88 have shown that there was no significant increase of the fuel failure fraction (Ref, 11) .

In the Russian Federation's experimental studies, major observations, besides kernel migration, were a vaporization-condensation effect to happen via radial microcracks in buffer and iPyC layers, and a fuel creeping effect caused by iso- lated gas-filled bubbles in the kernel (Ref. 187).

The goal of the fuel performance qualification program in China is to reach the FRG design values. No specific data for Chinese fuel have been given so far.

Available data on defective/failed TRIS4 particle fractions and heavy metal contamination from the different countries' fuel qualification program are summarized in table II .

For fuel particles with BISO coating, the end-of-life failure fraction to be used in German safety analysis calculations was derived from experimental data to be as high as 2* 10-3. A failure fraction of even 1 % was found under extremely conservative irradiation conditions and has been used as the design value for KFA safety analysis calculations for the THTR-300 (Ref. 51).

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 15 ä ü

d V

C a i C

cC ~n r'r ä, CD .. b ü M 00

Q ,f " r i s i H

0 \O V

Ca CL. n .. ü p ~ ~ ~ o rn

13 ~i yy sd

c~t Ö

° ~, ä © ä a ~ ~ w c ] c ar ö ,~ > > iet ü V ; m ME > _~ 1l

C] ~ A yC V p .~ `~ Z; p txo W -,Z C _y v% m Ü C rw ..w :; ü ~+ C,

2.2.1 Kernel Migration (Amoeba Effect)

Temperature gradients in coated particles due to extreme operating conditions or due to asymmetrical fuel kernel production lead to carbon transport from the "hot" side to the "cold" side of the inner pyrocarbon layer causing a migration of the fuel kernel toward the "hot" side (amoeba effect).

Kernel migration is not expected to occur in German HTGR fuel of a pebble bed reactor due to low power densities and a homogeneous fuel distribution .

Japanese HTTR fuel, however, is assumed to be exposed to severe temperature gradients. Experimental data on kernel migration distances from irradiation tests with high temperature gradients up to 150 K/cm have been gathered. From these measurements, a design equation for the kernel migration rate as function of the irradiation temperature has been derived demonstrating that even under severest HTTR conditions, the migration distance of the kernel calculated to be 55 gm was far below the safety design limit of 90 ,um, i.e. the sum of buffer and iPyC layer thicknesses (Ref. 11).

In former US modeling, kernel migration has been recognized to be important because of the expected large temperature gradients in large-sized HTGRs. The migration of oxide-based particle kernels with a failed coating was examined by heating them in a thermal gradient at temperatures of 1100-1350 "C. It was found that the kernel migration as a function of time, temperature and CO concentration was associated with the reduction of the oxide phases in the kernels (Ref. 52) .

Probabilities of SiC failure as function of the migration distance of the kernel toward the SiC layer are tabulated in a computer routine where the migration distance is dependent on time, temperature gradient, and kernel migration coefficients as function of temperature and kernel material . Data have been derived from experiments with HEU UC2 and Th02 kernels in BISO particles. However, failure by kernel migration is considered negligible for the present MHTGR core design.

A Russian calculation model has been developed (Ref. 14) which describes the kernel migration phenomenon in the temperature range 1000 - 1250 °C and in the burnup range 0.5 - 5 % FIMA. The basis of the model is the heat exchange from the surface of an unsymmetrical particle kernel with an external medium

2.0 HTGR FUEL, BEHAVIOR DURING NOR.1U1AL OPERATION 17 which induces a thermal gradient. The external medium consists of fission gases (Xe, Kr) and reaction products of oxygen (CO, CO2) which fill the free volume provided by the buffer and the gap between kernel and buffer. The equilibrium partial pressures of CO and CO2 are dependent on temperature and burnup. A diffusive transport of carbon (oxides) occurs from the hot to the cold side of the inner pyrocarbon layer. A comparison of model calculations with experimental results is not given in Ref. 14 .

Experimental studies in Russia have also been conducted at higher temperatures 1300-1800 °C and with temperature gradients up to 450-2500 Kjcm. Metallographic cuts through coated particles with c 12 % FIMA and a temperature gradient of 20 Kjcm at 1400 °C have revealed migration distances c 8 pm (Ref. 67).

2.2.2 Fission Product Interaction With Silicon Carbide

The interaction of the fission product species palladium with silicon carbide has been experimentally investigated at 3AERI for HTTR reference design fuel by measuring the depth of penetration of the resulting intermetallic compound into the SIC layer (Ref. 11). The maximum Pd-SiC interaction depth was found to be dependent on the cubic root of the calculated Pd amount released from the kernel. From these measurements, it could be concluded that with respect to HTTR conditions a maximum penetration depth of II Am at end-of-life could be expected which is less than half of the total SiC layer thickness. A quantitative study of the penetration mechanism has started.

In the US, reaction kinetics of the SIC attack by palladium have been experimentally investigated with the attack rate of fission products on the SiC layer found to be highly temperature-dependent (Ref. 53; based on experimental work by Montgomery) . Failure of the SIC layer is assumed to take place as soon as more than 50 % of the original SIC thickness has corroded. Although Pd is sus- pected to be a species of major importance, other metallic fission products are also believed to be involved .

2.2.3 Pressure Vessel Failure

According to the US modeling approach (Ref. 53), a pressure vessel model is regarded significant under normal operating conditions only for particles which do not meet the specifications (for instance too large kernels with too thin buffer layers) . Standard (intact) particles are not expected to fail . Particles with initially missing or defective buffer are assumed to fail completely after having reached 25 % of the design burnup while particles with missing or defective oPyC are assumed to show a more delayed effect with some temperature dependence. Both failure types produce exposed kernels. These assumptions for fuel performance during normal operation are incorporated into the calculation code SURVEY (Ref. 62) .

The effects of irradiation temperature, neutron fluence, burnup, particle component geometries and densities on the PyC and SiC layer stresses have been estimated using the computational results with analytical stress models. The model equation is based on the assumption of 40 % gas release from fissile kernels (20 from fertile kernels) and of no failure of the oPyC layer (Ref. 8) . Thermal creep reduces the potential of pressure vessel failure of the oPyC layer at high temperatures. For the US-MHTGR normal operation, no significant pressure vessel failure of standard particles is expected to occur (Ref. 8) . The US pressure vessel model is regarded more as a guide line for fuel particle design rather than a predictive tool for coated particle performance.

2.3 RADIONUCLIDE RELEASE DURING NORMAL OPERATING CONDITIONS

Fission product release behavior under normal operating conditions is of significant interest in safety analyses. Activity released into the primary circuit provides the major contribution to the accident-induced source terms especially in small-sized HTGRs, since the release from the fuel at elevated temperatures is expected to remain comparably small. Fission product activity is available either as coolant activity or as plateout activity on surfaces of the primary circuit or as dust-borne activity providing potentially significant exposure levels with a severe

2.0 HTGR FUEL BEHAVIOR DURI3N'G NORMAL OPERATION 19 impact on maintenance procedures. Within the active core, re-adsorption processes of metallic radionuclides from the coolant lead to an increased activity level in the outermost part of the fuel elements as it has been observed in the AVR reactor.

Table III lists, as an example, the fission product inventories expected within the primary circuit of the 200 MW(th) HTR-Module, following a 30-year normal operation. These inventories are subdivided into different groups which must be handled separately in source term estimations.

V C O

C O M O cd bA nrC 3 0 ö w

M F rr

C U w

E ä 0 0 0 w v 0 ö w ö ".rS-. v ö U Ü .2 y a. Cr_... 4r L U O U ca tM. v as 4-;4 0 1-i C U 0 ä a a. O . U L U i.. 0 0'a G.. p4 4 vs = CA ö G4 c.~ C.) O .O .=_ i~ . co.. . r Ü Ü C C -O t-' .G cd ö w F U U t7 `~ ~Ua

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATIONS 21 2.3.1 Metallic Fission Product Release

The transient release behavior of (long-lived) metallic fission products can be predicted by using diffusion models as described in further detail in section 3. An intact TRISO particle coating represents a highly efficient barrier against fission product release at normal operation temperatures (except for silver) . Therefore, the most important input data for coated particle performance during normal operation will be the fractions of defective/failed coated particles and the fraction of heavy metal contamination in the fuel element graphite in combination with the transport data in kernel material and fuel element graphite . In addition to diffusion, other release mechanisms during normal operation are the recoil effect and the knockout effect. Both represent a geometrical problem and are not dependent on temperature making them relatively more significant at lower temperatures.

Gaseous precursor nuclides of metallic fission products must also be considered, in particular the strontium isotope Sr-89 . It is not clearly known to what extent the presence of the nuclides Cs-137 and Sr-90 in the coolant is caused by precursors, by formation of contaminated dust due to fuel element abrasion, or by direct release of the radionuclides and/or their compounds. With respect to the Fort St. Vrain HTGR, diffusion tube measurements have indicated that direct release does not occur. Sources of primary circuit contamination are the formation of contaminated dust and gaseous precursors only (Ref. 84). In contrast, direct release seems to be dominant in the AVR reactor (Ref. 60) and is therefore assumed in actual German safety analyses . Sources of this direct release by coated particles with intact or defective coatings and by fuel contamination of graphite are considered separately. Reliable data on fuel in particles with defective coatings and in graphite are obtained by the fuel element qualification program (see previous sections).

The pebble version of the KFA diffusion model FRESCO (Ref. 30) (see also section 3.2) has been widely used to describe the metallic fission product release behavior from a spherical fuel element both under irradiation and elevated temperature conditions. This model includes specific irradiation effects such as recoil and the buildup of fission product inventories dependent on the decay constant. It is based on effective diffusion coefficients for the fission product species in the different fuel materials. Transport data for normal operating conditions are given in Appendix A. By choosing average fuel operating conditions, a FRESCO-11

calculation can be taken to describe a representative fuel element during its lifetime even for a fuel sphere to pass several times the active core (multipass loading scheme) as planned for the German small-sized HTGRs. The release results can then be extrapolated for all fuel spheres over the plant's lifetime.

The JAERI computer code FORNAX (Ref. 35) is similar to the diffusion code FRESCO-Il in describing the metallic fission product release from the particle kernel by diffusion and recoil and the diffusive transport through the coating materials in a fuel element. Three different types of fuel particles are considered in the FORNAX model: standard (intact) particles, failed particles (= exposed kernels) and - at a stage in between - particles with a degraded SiC layer simulated by a larger diffusion coefficient. Calculated results show good agreement with residual cesium and silver activities in the particles and plated-out activities inside the capsules from irradiation tests (Ref. 35) . The FORNAX model is also applied to accident conditions.

The KFA diffusion code SPTRAN (Ref. 55) has been recently modified for use. under normal operating conditions (Ref. 17) . No significant difference in the method of modeling fission product release from the fuel between this code and FRESCO-Il can be detected. Validation calculations of irradiation tests or reactor experiments have not yet been presented for SPTRAN.

The GA codes FIPER (Ref. 151) and TRAFIC (Ref. 65) determine the release of metallic fission products from the HTGR core into the primary coolant circuit. Both codes model a one-dimensional Fickian migration through the fuel particle, fuel compact, and structural graphite to the coolant hole surface. Sorption on fuel compact matrix material and evaporation from the graphite surfaces into the coolant are taken into account as well as the recoil effect and radioactive decay. An extended version transforming FIPER into a multiple-path, multiple-species code, called TRAMP (Ref. 152), includes diffusion as a result of thermal gradients, the in-grain and grain surface diffusion as a result of concentration gradients, and the convection of the coolant through the permeable graphite (pores). The TRAFIC code is specialized for a rapid computation of block-averaged metallic fission product release at many spatial positions in the HTGR core using an irregular space and time dependent temperature history.

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 23

2.3.2 Uptake of Fission Product Metals by SiC

The fission metal content of SiC following the irradiation of TRISO-coated fuel particles in the High Flux Isotope Reactor (HFIR) at ORNL, the R2 Reactor at Studsvik and the Fort St. Vrain (FSV) Reactor at temperatures between 512 and 1200 °C, burnups between 0.9 and 45 % FIMA, fast neutron fluences between 1 .9* 1025 and 7.$* 1025 m-2 and for 171 to 519 effective full power days (efpd) has been measured (Ref. 32). Analysis of these measurements has lead to the con- clusion that the quantities of the metallic fission products, cesium, cerium, europium, ruthenium, and antimony, are proportional to the number of neutrons that have passed through the SiC. The linearity of the relation between the quantity of fission products in the SiC and the neutron fluence is improved by accounting for the differences in the ratio of the number of atoms in the SiC and the number of neutrons that have passed through it. This ratio is 3 to 4 times larger for the graphite moderated reactor, FSV, than for the water moderated reactors, HFIR and R2.

No explicit dependence of the atom content of the SiC was found on time, burnup, temperature, or kernel enrichment. A qualitative model to account for this was conjectured to have the following components:

l . Structural damage in the SiC coating from the slowing down of fast neutrons creates sites at which the fission products bind or paths by which they may enter the coating and

2. the fission product atoms move into the SiC coating to occupy these sites and perhaps along the paths generated by fast neutrons .

According to the first element of the model, the fission product quantities are proportional to the number of sites created during the slowing down of fast neutrons. If the fission product content of the SiC is proportional to the number of damage sites in the irradiated SiC, then an estimate of the probability of neutron damage with respect to fission product binding can be made. From an alternative viewpoint, the effect of the neutron damage is to increase the solubility of fission products in SiC. The strengths of the binding of fission products to SiC appear to be larger than small multiplies of kT where T is in the temperature range 512 to 1200 °C.

2.3.3 Fission Gas Release

The activity of gaseous fission products in the coolant is a direct indicator of fuel performance. Its main sources are the heavy metal contamination in the fuel graphite and defective/failed coated particles. Standard particles within the specification limits are not expected to contribute to the release of fission products under normal operating conditions .

For the release of long-lived fission gases from the kernel, the classical formu- lation for diffusive release from a sphere is applied, with the fractional release given by (Ref. 57):

where t : irradiation time [s] D' : reduced diffusion coefficient [s -l] D' _ D/a2 D : diffusion coefficient [m2/s] a : kernel radius [m]

For short-lived gaseous fission product species, production and release from the kernel will quickly reach an equilibrium state. The Booth model (Ref. 15) is most commonly taken to determine release rate to birth rate ratio (R/B) as function of the diffusion coefficient in the kernel and the decay constant. The Booth model is applicable to the release of gaseous fission products both from defective/failed coated particles, i.e. exposed kernels, and from matrix graphite grains (which contain the heavy metal contamination) as simulated by equivalent spheres . It can also be used for gaseous precursor nuclides of radiologically relevant long-lived metallic fission products (Xe-137, Kr-$9, Kr-90). The analytical formula of the Booth model is:

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 25

3 R / B = x " ( coth x --- x ) (2) where

2 ~. . a x D

and d : decay constant [s -i] a: kernel radius [m] D/a2: reduced diffusion coefficient [s ^1]

If x > 1) then:

R B ,., 3 - 3 . D / z 2 (3) , . a

showing the characteristic feature of the 1 I-,5._ dependence for short-lived isotopes (valid for a smooth surface sphere) .

The HRR model for fission gas release from defective fuel particles (Ref. 56) is also based on Booth's R/B equation (2) and its approximation . But this model distinguishes between the different components, grains and pores, of both the particle kernel and the buffer layer. The complete equation is given by:

° - (A, T) F - E.fkk 9 ( + fkp + fpk' ( + fpp 1 ' ( ) (4) B )kk B )pk T p

where fkk : ratio of birth rate in particle kernel grains over B fkp : ratio of birth rate in particle kernel pores over B fpk : ratio of birth rate in buffer grains over B fpp : ratio of birth rate in buffer pores over B f i =bi/B and Zbi =B F : factor describing the type of defect of the particle coating (F = J. for the most releasing type : kernel + buffer only) (R/B) p : release-to-birth ratio for the pores

The magnitude of f-values which contain the recoil effect are empirically estimated. (R/B)p is supposed to show the T3/2 dependence of a gas phase diffusion but due to a negligible retention of the pores, it can be approximated by (R/B)p 1 . At lower temperatures c 700 °C, the diffusion of gaseous atoms from recoil sites in the buffer is more significant than the diffusive release from the kernel grains.

The US model for steady-state fission gas release from exposed kernels also begins . with the Booth model approximation, then adds empirical correction terms for temperature and burnup (Ref. 57). For radionuclide i and element j:

3 " .f(T") " ,f(Bu)

where Q : activation energy [J/mol] R : ideal gas constant, R = 8 .3143 [J/(mol K)] T : temperature [K] Bu : burnup [q FIMA] A, To , c, n : constants

For each gaseous fission product element (Kr, Xe) and for each fuel type (UO2) UC2, UCO, ThO2), the values of D, A, Q, To, Q, and n must be determined.

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 27

For transient release of short-lived fission gases from bare kernels during postirradiation heating, the US uses a diffusion-trapping model with the fractional release given by (Ref. 73) :

e_a( l _ .fß( 1 7 { + S " r (6)

where t : time [s] fR , a, S : parameters to be determined

The isothermal heating experiments typically result in fractional release curves which initially increase rapidly, then more slowly with a linear increase with time. For this case, the empirical parameters represent the following:

f9: magnitude of the initial rise in the curve tt : rate of the initial rise in the curve S : slope of the slowly rising, linear portion

The CEGB model for describing rare gas fission product release from UO2 (Ref. 58) is represented by the approximated Booth formula:

RIB W v .

where S/V is the surface to volume ratio of single UO2 crystals (magnitude c-- 11 mm-1) . RjB experimental data from irradiation experiments have been interpreted such that the deduced diffusion coefficient is composed of three rate- controlling terms, one representing high-temperature intrinsic diffusive behavior and the two others controlled by irradiation damage processes. A time dependence of the effective diffusion coefficient was found based on the trapping and re-solution of the gaseous atoms at intergranular bubbles . The two athermal irradiation induced mechanisms dominant at low temperatures are the recoil effect and the knock-out of atoms close to the surface.

The JAERI approach of modeling the release of short-lived noble gases from failed coated particles and matrix contamination (Ref. 44) is based on an empir-

ical equation which has been generated from previous irradiation experiments with fuel compacts and loose particles. It describes the Kr-88 RIB measurements per failed particle as a function of the irradiation temperature. The release of fission gases other than Kr-88 is then determined by assuming constant ratios of the RIB to that of Kr-88 with the ratios determined using an analytical diffusion model which also considers the effect of precursors:

(R/R)i (RIB)K,.- 88

where Ki : parameter for nuclide i

Good agreement except at lower temperatures was found when applying this model to irradiation tests with fuel compacts which contained artificially failed particles . However, R/B measurements from compacts with contaminated matrix were different from the calculated release behavior predicted with this model. The model is taken to estimate the activity circulating in the primary system of the HTTR.

Another JAERI model to calculate fission gas release from oxide fuel is based on a modeling approach developed at Chalk River/Canada (Ref. 39) . It is an extended Booth type model which takes account of a diffusive transport in the oxide fuel, a partial retention of gas in trapping sites which are available as either closed voids or larger irradiation-induced crystal defects, and a possible re- solution of trapped gas with a subsequent diffusive transport. This model is similar to the diffusion-trapping model of transport in graphite.

A numerical determination of the gas release during normal operation is given by the KFA computer code STADIF-II (Ref. 18) . It describes the steady state fission gas and iodine release from defective particles, recoil effect, and graphite contamination by using an uncoupled two-phase (grain, pore or grain boundary) diffusion model, but considers no sorption effect on graphite surfaces. The code also allows for calculation of the coolant activities taking account of the purification system, for iodine in addition plateout constants . Calculated krypton and xenon release values are in good agreement with AVR experimental results.

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 29

The US code RAD (Ref. 83) determines the RJB values of various gaseous fission product isotopes and thus can estimate their inventories in the fuel, in the primary coolant, and on the surfaces exposed to the coolant within the HTGR system. The main input data are the diffusion coefficients derived from experiments with as-manufactured fuel compacts and defective fuel particles, and the removal rates due to plateout and due to the purification system.

The Russian method3 of determining the fission gas and iodine release under normal operating conditions is characterized by a temperature dependent leakage velocity R(T) - equivalent to a change of the activity per unit time - of the nuclides out of a core volume element AV (Ref. 9):

R(T) - AV(T)

AV(T) corresponds to the part of fuel elemdnts in the core at temperature T.

The main sources of fission products are identified as defective particles (0) and heavy metal contamination of the particle coatings and the matrix graphite (co). Regarding - for completeness - also particles with an intact coating (mt), the overall leakage velocity reads as follows:

R = RO + Rw -i- R,n~

According to the so-called two-group activation model (Ref. 189, see also section 3.3), the leakage velocity Ri is given by:

Rj _- mr - .f (ag, z) + ( 1 MI) - f(am, -c)

s The following explanation of the Russian model as described in Ref. 9 reveals some inconsist- encies and needs further discussions among the experts.

where

a .T I - e` 1 Pah-T) = 1 I _ e-A " ? aj + A

and leakage constant [s -1] indices j=g and j=M represent the two groups of fission products considered in the activation model A. : decay constant [s-1I z : irradiation time [s] mi : spectral characteristics [Bq/s] derived from experiments with irradiated fuel elements index i = 0 : defective particles i = cu : heavy metal contamination i = mt : intact particles

For most of the volatile nuclides, the approximation ag > .? > aM is valid, thus:

R(T) = C ~ (7) " 0 " B I " AV(T)

E m. (7) ' w . B " AV(T) (9) E m»st (2) " ( 1 - ~ß .- ro ) " B " AV(T)

where B : birth rate ¢ : fraction of defective particles co : fraction of heavy metal, contamination mt : fraction of intact particles mt=1 -0-w

Integration over the total core volume provides the overall volatile fission product release into the coolant. Leakage rates from defective particles and from heavy metal contamination have been measured in experiments.

The principal disadvantage of existing release models for normal operating conditions is their inability to handle all the important phenomena in the active core of an HTGR at the same time. For example, the normal operation release results calculated using the diffusion model FRESCO-I1 for a single spherical fuel element must be extrapolated over the whole pebble bed. The feed-back given by

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 31 gas-borne activity on the release behavior which has been observed in AVR fuel elements to create an increased metallic fission product concentration in the outermost fuel-free graphite shell due to re-adsorption processes in colder core regions, is not considered in the models. Modeling of the coolant inlet effects also requires a connection to plateout processes in the primary circuit. Consideration of this readsorption effect as well as the production of dust by fuel element abrasion and its effect on fission product transport, and of the effect of precursor nuclides, would complete a comprehensive model for the core release behavior under normal operating conditions .

2.3 .4 Fission Product Activity Distribution in the HTGR Primary Circuit at Accident Initiation

In the AVR, continuous measurements of noble gas activity were conducted (Ref. 60). Particle failures could then be detected by a considerable increase of the activity. The VAMPYR-I experiment was designed to collect solid fission products on filters as dust-borne or free (atomic or molecular) activities in the AVR coolant gas. Their activities are several orders of magnitude lower compared to the gaseous species. The measured activity in the coolant mostly originated from the heavy metal contamination in the fuel matrix. Cesium profiles in fuel elements with normal particle performance show an increase of concentration near the surface due to adsorption of cesium from the gaseous phase in cooler parts of the core (cross-contamination) . The reduction in this adsorbed activity observed over the years reflects the gradual replacement of BISO by high quality TRISO fuel within the AVR core.

The HRB fission gas release model was used for the licensing procedure of the THTR-300 and has been recently applied to a comparison with the measured coolant gas activity after 423 efpd of THTR-300 operation (Ref. 59). Activities of various noble gases have been measured quasi-continuously. Considering the R/B values as a function of the decay constant, it could be concluded that, in accordance with the design model, the dominant source of gaseous activity is due to the uranium contamination of the matrix graphite from the manufacturing process. The activity increase of a factor of 2 after about 100 efpd could be explained by direct recoil from exposed fuel in mechanically damaged spherical fuel

elements. Nevertheless, the activity was not higher than 4 % of the design limit 104 of 3* for heavy metal contamination and 2* 14-3 for in-service failure fraction.

For the Fort St. Vrain HTGR, a comparison of predicted with measured fission product release has been made (Ref. 61) which demonstrates that the calculation model overpredicts the measurements by a factor of 5.4 The dominant source for coolant activity was the heavy metal contamination. The release of fission metals is insignificant compared to the release and subsequent decay of their gaseous precursors.

Plateout distribution of fission products in the primary circuit under normal operating conditions is calculated at KFA with the codes SPATRA or PATRAS which combine ad-/desorption behavior and mass transfer, as it was first proposed for HTGR plateout calculations by Kress and Neill (Ref. 87). In the SPATRA code (Ref. 85), the ad-/desorption equilibrium partial pressure on metals is approximated by:

QHd,, = " e R " T " (14) pi la o'0

where pi : partial pressure of nuclide i [Pal a : concentration of fission products with respect to the 2 geometrical surface [mol/m 1 ao : monolayer concentration on geometrical surface ao N 1'10 -5 [mol/m2] desorption enthalpies in the sub monolayer regime [J/mol) Qkides : R : gas constant, R = 8 .311+3 [J/(mol K)] T : temperature [K]

This equation is easily obtained by equating the collision rate of nuclides onto the geometrical surface with the desorption rate. The latter is proportional to the lattice vibration frequency and to an Arrhenius factor containing the desorption

4 The overprediction was primarily the result of including a hydrolysis factor of 6.3 when the effects of hydrolysis did not persist long enough for this factor to be relevant. Furthermore, the factor, in view of what is now known, is too large.

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 33

enthalpies. As shown in Ref. 150 in more detail, the assumption of an ad- /desorption equilibrium is a reasonable approach .

Introducing equation (10) into the mass transfer equation

dn ß ~ Cgas - Cgr) (11) F ' dt '

where F : area [m21 n : number of moles mass transfer coefficient [m/ sl concentration in coolant egas : [m_31 31 agr : concentration in graphite [m_

leads to:

8 AHdes Q (Pi,BULK - 10 " e- R " T (12) d t R ~, T ` 60

where da/dt : deposition rate [mol/(m2 s)j

which is solved (in a more complicated form) in SPATRA. Equation (12) is valid only for a low fission product burden on surfaces which is usually true for the normal operating reactor. In case of higher surface concentrations saturation effects must be taken into account, meaning that fission product concentrations cannot be considered independently from their adsorption behavior. Coverage of effective surfaces approaching a monolayer also leads to a significant decrease of desorption enthalpies. However, due to surface roughness, effective surfaces in HTGRs may be at least one order of magnitude larger than the geometrical surfaces.

Values of desorption enthalpies are in the range 220-260 kJ/mol for cesium and silver and 110-180 kJimol for iodine, depending on the adsorbent and its surface conditions . Cesium desorption enthalpies seem to be larger on oxidized surfaces than on clean metallic surfaces, whereas iodine desorption enthalpies are especially small on oxidized surfaces. A quantitative relation between surface condi-

tions and desorption enthalpies, however, is not available up to now.s One should keep in mind that the desorption enthalpy is part of the exponent in equation (12), thus the data range of AHdes(Cs) given above translates into sorption isotherm differences of several orders of magnitude.

Strontium plateout behavior is less well known. Thermochemical estimations have indicated that strontium in the presence of oxide layers may be converted into SrO or into another oxidic phase. In that case, equation (1Q) is not applicable, and the usual vapor pressure equations for pure compounds must be assumed. Vapor pressures of oxidic strontium compounds, however, are extremely low. Pure compound vapor pressure equations may also be applied in case of iodine plateout under conditions of solid Fe12 formation, i.e . for less oxidized metal surfaces at low temperatures (Ref. 88). Because of their relevance for source term estimations in water ingress accidents (see section 4.4), the knowledge concerning desorption and vaporization enthalpies under reactor conditions must be improved by additional experiments.

SPATRA postcalculations (Ref. 85) of experimental plateout profiles of Ag-11Qm for the in pile experiment VAMPYR-11 in the AVR-reactor (Ref. 89) have shown that in the high-temperature part of the test section, plateout is controlled by the ad-/desorption equilibrium, whereas in the low-temperature region mass transfer is dominant. In the same experiment, cesium plateout seems to be strongly influenced by graphitic dust adhered on metallic surfaces, resulting in a nearly flat plateout profile. This effect cannot be reproduced with the code SPATRA up to now. However, the extremly high dust content within the primary circuit is probably an AVR-specific problem and may not be generally applicable to modern HTGRs.

In the VAMPYR-I experiment, the influence of dust on the cesium plateout curves does not appear as strong as that observed in VAMPYR-II (Ref. 89). The adsorbent metals at least in the high-temperature part of this experiment, however, are not representative for the HTGR primary circuit. In addition, consid-

S However, a qualitative explanation might be provided by an atom's electronegativity (affinity for acquiring electrons) and oxidation state. Because metals (Cs, Sr, etc.) readily give up electrons and oxygen prefers electrons, they are compatible. But iodine and oxygen would compete for electrons and thus be less compatible. This interpretation is consistent with the observed desorption enthalpies.

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 35

erable activation processes take place in the high temperature part of VAMPYR-I influencing in particular the Ag-110m concentrations. Flow conditions in this loop differ significantly from those in VAMPYR-1I. Iodine plateout in the VAMPYR-I loop is obviously controlled by ad-/desorption equilibrium and mass transfer only (Ref. 89) and can therefore be easily handled by SPATRA .

SPATRA also contains a simplified model for consideration of absorption processes. A partition coefficient is assumed between adsorbed and absorbed state. The transport into the adsorbing material is regarded as simple Fickian diffusion. Absorption effects are experimentally confirmed to some extent for cesium (Ref. 93); they are, however, not well understood up to now. Mechanisms such as volume diffusion, diffusion in the micropore or grain boundary system of the oxide layers and growing of oxide layers must be taken into consideration. There are experimental arguments for absorption: diffusion profiles of cesium have been found within the oxide layers of metals for a depth of some porn (Refs. 94, 95, 96) at temperatures of 600-800 °C for diffusion times on the order of hundreds to thousands of hours. In addition, leaching with water removes only on the order of ten percent of the plated-out cesium (Refs. 97, 93, 95), especially in the high- temperature deposition area. Plateout of cesium sometimes shows a remarkably small dependence on temperature which has often been explained by (temperature independent) absorption.

In many cases, however, mass transfer may have diminshed the temperature dependence. Simple absorption or penetration mechanisms are in conflict with the following fact: at deposition temperatures > 400 °C, the water leach yield does not show a clear temperature dependence. If adsorption decreases in favor of absorption with increasing temperatures, as assumed in simple absorption models, a remarkable decrease of the leach yield must be expected. In addition, diffusion profiles have been measured also at low temperatures (C 150 °C) and were found both in mass transfer controlled plateout regions and in ad-/desorption controlled regions, the latter sometimes showing no significant deviations from simple ad-/desorption profiles. This might be explained by cesium diffusion along the inner oxide surfaces/grain boundaries with a rate which is fast compared to the rate constants of cesium desorption into the gas phase. From a physical point of view this is possible, because in surface diffusion, cesium atoms do not completely leave the attraction range of the surface and the corresponding activation energies of diffusion are much smaller than desorption energies. This explanation

is compatible with the low leach yield, because penetration of water into the oxide layer micropore system is strongly impeded.

Besides diffusion within the micropore system of the oxide layer, some real absorption (i.e. irreversibility relative to ad-/desorption equilibrium) with inclusion of cesium into the metal/oxide volume may also occur. Although absorption hinders remobilization of plated-out fission products, it is conservatively not considered in safety analyses up to now because of the uncertainties. An overview of models concerning absorption is found in Ref. 63.

Calculations on the basis of equation (12) for the primary circuit of actual HTGRs (Peach Bottom (Ref. 90), AVR (Ref. 91)) using the US codes PAD/PADLOC (Ref. 92), indicate a sufficient agreement with the measured plateout distribution for Cs-137 and Sr-90 in these reactors . This agreement may be taken as a partial validation of the physical plateout model which is used in a nearly equivalent manner in PAD/PADLOC and SPATRA.

As mentioned above, the KFA code PATRAS (Ref. 86) is similar to SPATRA with respect to adsorption but considers a temperature independent penetration' coefficient (which is identical to the absorption probability of a gas atom colliding with the surface) and a re-evaporation of the absorbed nuclides as soon as solubility limits are reached. Some experimental facts conflicting with a simple absorption/penetration model are given above. In addition, from a physical point of view, it is difficult to explain the absorption probability of a gas atom colliding with a surface to be independent of temperature (or kinetic energy of the gas atom) and rather depends for a given nuclide on some (temperature independent) material propertiers only. All transport processes into the volume (volume diffusion, grain boundary or micropore diffusion) are coupled to a significant activation energy which necessarily requires a temperature dependence of that penetration coefficient. Although the penetration coefficient is an empirical factor without a definite physical meaning, it can be shown that this model provides reasonable results for plateout problems, in particular for low fission product concentrations (where the above mentioned solubility limits do not play any role) .

The plateout codes SPATRA and PLATO (Ref. 98) also contain the penetration model as an option . An improvement of the penetration model of PATRAS (and the other codes) should contain the option of a temperature dependent penetration coefficient.

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 37

Input for the calculation models for estimation of the fission product distribution at accident intitiation are basically taken from best estimate data of irradiation experiments with fuel elements (Ref. 196), of reactor experience (Refs. 60, 89, 90, 99), and, in case of plateout, of laboratory experiments (Refs. 93, 97, 100) . Ad- ditional data are expected from a research program at FZ Rossendorf on fission product release in HTGR water ingress accidents, where plateout (and remobilization) of Cs and iodine will be measured in a thermogradient tube and in a large plateout loop (Ref. 101) . A major goal of this work will be the examination of the dependence of plateout parameters on surface conditions.

While in safety analyses best estimate values of desorption enthalpies are used to estimate the plateout distribution, their lower boundaries are taken to calculate the equilibrium cooling gas activities. This procedure leads to lower limits for the relative plateout per pass and to upper limits of the atomic or molecular cooling gas activities. Since the amount of plateout for iodine and metallic fission products is always more significant than the atomic/molecular gas-borne activity, the different handling of gas-borne and plateout activity does not lead to conflicting mass balances. The equilibrium activity of noble gases must be calculated by considering sinks like primary circuit leakages and gas purification systems.

With respect to the empirical modeling of dust (see section 5.2), the uncertainties concerning its production, settlement and interaction with gas-borne or plateout activities are high. The radionuclide activity values given in table III at the beginning of this section are upper limits of the uncertainty scatter which may be reduced with increasing knowledge. In connection with dust contributions to the equilibrium cooling gas activities, it was found that an equilibrium like dust concentration in the coolant is obtained only for uniform flow conditions (see section 5.2). Based on experimental evidence from the AVR, a fraction of 10-7 of the total mass of dust within the primary circuit is assumed to be gas-borne under equilibrium conditions. For the 200 MW(th) HTR-Module, this fraction will -9 contain about 5* 10 of the Cs-137 inventory released from core into the primary circuit (or 5*10" 13 of the overall reactor inventory).

Conclusions from table III are that for the HTR-Module :

" less than 2* 10-4 of the inventory is found outside intact coated particles,

+ the equilibrium activity within the coolant is extremely low due to low release from the fuel and due to plateout on primary circuit component surfaces,

the inventories of defective particles and the plateout on the metallic primary circuit surfaces represents most of the activity outside intact coated particles.

2.0 HTGR FUEL BEHAVIOR DURING NORMAL OPERATION 39 40

3.0 HTGR FUEL BEHAVIOR DURING CORE HEATUF ACCIDENTS

For modern HTGR TRISO coated fuel particles, the SiC layer is expected to be a highly efficient barrier against radionuclide release. For the fission gases, the pyrocarbon layers also show good retention capabilities.

Increasing fuel temperatures lead to a higher coating failure fraction and to enhanced permeability of the coating layers for fission products. The release behavior of fuel particles with no intact coatings is determined by the retention capability of the fuel kernels.

3.1 THERMODYNAMICAL BOUNDARY CONDITIONS

Maximum fuel temperatures under core heatup accident conditions in a depressurized reactor are strongly dependent on the thermal power and power density of the plant. Medium-sized HTGRs such as the German designs THTR-300 or . HTR-500 would experience fuel temperatures up to about 2360 °C (depressurization within 150 h, Ref. 79) and 2550 °C (Ref. 80), respectively, under these conditions. The temperature range of interest, however, has been shifted towards lower values in the last decade with the HTGR community looking in detail at small modular HTGR units where fuel temperatures do not exceed 1600 ° C even under severest accident conditions. For the German 200 MW(th) HTR-Module, the maximum fuel temperature is predicted to be 1550 °C (Ref. 81). In the 450 MW(th) US-MHTGR with an annular core, a maximum temperature of about 1500 °C is reached under core conduction cooldown conditions (Ref. 4) . The corresponding value used in predictions for the Japanese 30 MW(th) HTTR is c 1600 °C (Ref. 34) .

Helium coolant velocities after loss of forced convection conditions are expected to be reduced down to the order of cm/s, thus increasing the capability of graphitic and metallic surfaces in the core cavern and in the primary circuit to reduce metallic fission product activities by sorption processes (see section 3.4).

In core heatup simulation experiments, the heating temperature conditions have been defined on the basis of the above mentioned predictions. Typical transients used to simulate medium-sized HTGRs are ramp tests with 2500 °C to be

3.0 HTGR FUEL BEHAVIOR DURING CORE HEATUP ACCIDENTS 41 reached within 8 h, 30 h, or 80 h while for small-sized HTGRs, isothermal heating tests in the range 1500-1800 °C are preferred (Ref. 120) .

3.2 COATED PARTICLE PERFORMANCE UNDER ACCIDENT CONDITIONS

The mechanical failure of a particle coating has always been a basic subject of investigation comprising the analysis of stresses imposed on the coating due to internal fission gas pressure, the dimensional changes of pyrocarbon (especially in BISO fuel particles), and SiC degradation processes in TRISO fuel particles . The theoretical analysis was accompanied by the development of experimental methods to measure SiC strength and its statistical distribution.

A first step of modeling particle failure was to evaluate irradiation experiments and heating tests at elevated temperatures in order to find the lower and upper limits at which to expect zero failure or complete failure, respectively. Early US particle failure diagrams have defined failure isopleths in a burnup - fast fluence diagram under normal operating conditions (Ref. 69) or regions for "no coating failures", "partial failures", and "100 % coating failures" as a function of irradiation time and fuel temperature under accident conditions (Ref. 68). An- other simple approach was developed for TRISO particles in the early 1980s by Goodin (Ref. 19), who derived from experimental data an exponential particle failure function dependent on fuel temperature or in a more complicated but still empirical way by adding a time dependence (Ref. 20) .

Information about the failure of particle coatings under elevated temperature conditions has been gained from series of heating tests . These data have been used to develop corresponding calculational models for both reproducing the experimental data and predicting coated particle performance in future HTGR designs.

Several mechanisms have been found to be responsible for particle failure under core heatup accident conditions:

1 . Pressure vessel failure of standard (intact) particles or in particles with defective or missing single coatings

2. Failure of the SiC layer by fission product corrosive attack on silicon carbide

3. Failure of the SiC layer by thermal decomposition of the silicon carbide

The KFA code PANAMA-1 (Ref; 21) determines the fraction of failed TRISO particles under accident conditions by combining two different failure mechanisms. The first one is a pressure vessel model which is based on the physical description of the spherical SiC layer to act as a pressure vessel. The SiC layer is expected to fail as soon as the stress on the coating caused by the internal gas pressure has exceeded its tensile strength . The gas pressure inside the particle, as derived from the ideal gas law, is strongly dependent on temperature, burnup, fission yield of stable gases, fission gas release from the particle kernel, and oxygen release by fission resulting in CO formation . The stress on the coating is additionally increased by a weakening of the SiC layer due to corrosive attack on the inner surface by fission products. The SiC corrosion is modeled by a thinning rate.6

The tensile strength is a material characteristic of the silicon carbide which has been measured in independent SiC ring crack tests (Ref. 23). This strength is considered to be reduced by fast neutron fluence. The oPyC layer which is pre- sumed to contribute to the tensile strength is conservatively neglected in this model.

The second failure mechanism, thermal decomposition of the silicon carbide, be- comes dominant at higher temperatures between 1600 and 2000 °C depending on heating time. This has been described by Weibull statistics with pre- exponential factors and Weibull moduli empirically derived for unbonded particles and for particles embedded in a fuel sphere. A thermal decomposition thinning rate has been derived by Benz from SiC weight loss measurements on coated particles without oPyC after heating (Ref. 24). A PANAMA-1 calculation is assumed to result in a fraction of particles with a simultaneous failure of all coatings (resulting in exposed kernels) and the fission gas release, respectively,

however, it might be not 6 A uniform thinning of the SiC layer is an assumption in the model; conservative since local thinning has been observed in Japanese experiments.

3.0 HTGR FUEL BEHAVIOR DURING CORE HEATUP ACCIDENTS 43 from the failed particles. PANAMA-I uses this as a conservative assumption rather than modeling the failure of individual coatings .

The experience gained with the PANAMA-1 model so far has demonstrated in most cases good agreement with measurements from German heating experiments covering both observed krypton release characteristics of sudden bursts (complete coating failure) and of a gradual increase typical of diffusive transport through the still intact oPyC layer (Ref. 12) . Fractions of gas release from fuel spheres which were exposed to extreme irradiation conditions in MTRs are in some cases overpredicted by PANAMA-I up to several orders of magnitude.

A joint US/FRG modeling effort resulted in the so-called "Integrated Failure and Release Model for Standard Particles" as a new definition of particle failure . The cesium release is here supposed to indicate a failure of the SiC layer. The diffusion of cesium is assumed to be negligible in this model under the conditions to which the model is applied . The observed delayed krypton release profile can be explained by assuming a diffusive transport through the still intact oPyC layer whereas there is no holdup for cesium in pyrocarbon at elevated temperatures. This is due to the observation of almost identical release curves for cesium and krypton from particles which did not have an oPyC layer (Ref. 28).

In the original version of this model from 1985 (Ref. 28), both thermal decomposition and corrosion of the SiC layer were taken into account as degradation mechanisms for the SiC layer. Similar to PANAMA-I, the experimental data from Montgomery and Benz, transferred into thinning rates, have been incorporated into a statistical distribution of the SiC degradation process. Its convo- lution with the distribution of SiC layer thicknesses leads to a complicated formula which is approximated by a Weibull distribution for the probability for a SiC coating to fail . Corresponding Weibull moduli and pre-exponential factors were derived from a small number of German and US heating tests up to 2500 °C.

The revised version of the "Integrated Failure and Release Model for Standard Particles" from 1988 regards thermal decomposition as the only SIC degradation process (Ref. 8). The Weibull modulus and pre-exponential factor were derived from an extended set of German heating test data. An activation energy was derived from these data which was very close to Benz' value for thermal decomposition of the SiC layer, therefore it was concluded that SiC corrosion was in-

significant for these heating tests. The pre-exponential factor contains a strong dependence on irradiation temperature and burnup and a weak dependence on fast fluence, which was the result of the fitting procedure rather than any physical interpretation. The particle failure fraction in the 1988 version has a minimum value of b* 10-6 as cutoff limit to agree with observations from low exposure AVR fuel under heating conditions.

These models were incorporated into the KFA codes PANAMA-II (original version from 1985) and PANAMA-III (revised version from 1988) and into the GA code SORS . 7 Both KFA and GA codes produce basically the same results (Ref. 8) . Due to the larger experimental data base, the application of the revised version provides better agreement with the measurements than the original version of the statistical model. However, for many heating tests and probably for most of the small-sized HTGR core regions under conduction cooldown conditions, the minimum for the SiC failure rate given in the model equation will not be reached which makes this cutoff limit rate-determining and not the irradiation conditions .

The JAERI model of simulating TRISO particle failure uses an approach similar to PANAMA-I with the SiC layer representing a pressure vessel (Ref. 25). The internal gas pressure is based on the Virial equations (expansion of the equation of state in terms of powers of T) describing non-ideal gases. In addition to fission gases, also alkaline earths, halogen, and tellurium species are expected to contribute to the pressure buildup in the particle. At the same time, fission-induced release of oxygen from UO2 results in a large CO partial pressure. Free volume to take up the fission gases is made available by the porosity in the U02 kernel and in the buffer layer. The oxygen-to-uranium ratio which is dependent on the burnup, is estimated from thermodynamics resulting in a lower CO pressure than the theoretically maximum one. The stress induced by the internal gas pressure is compared with the SiC strength which follows Weibull statistics and failure occurs when they are equal. SiC strength is reduced by the porosity which is caused by thermal decomposition of the layer and is preferentially developed at grain boundaries of the SiC. A 4 % porosity in the SiC layer is considered to make it permeable to metallic fission products. The model includes the statistical variation of the normally-distributed particle fabrication parameters by adopting a random sampling method. SiC strength and thermal decomposition (or

The actual SOBS code is always referenced by its original publication in 1974 (Ref. 27), though there are later developments in the calculational models now being implemented in this version.

3.0 HTGR FUEL BEHAVIOR DURING CORE HEATUP ACCIDE\TS 45 porosity increase) activation energy data for Japanese fuel particles have been derived from heating tests with both irradiated and unirradiated coated particles. These data exhibit much higher values than corresponding data for German and US fuel (see Appendix B) which should be the subject of further detailed comparative analyses. Postheating examinations revealed a significant delay from the SiC failure to the total coating failure. Plastic deformation of oPyC layer after SiC failure results in a "ballooning" of the particle which reduces the inner gas pressure buildup, but makes the oPyC permeable for fission gases at large stresses.

A modeling approach for a pressure vessel without adding a corrosion mechanism was made in the GA code CONSTA (Ref. 26). Calculational results with CONSTA were compared with experimental results with varying success. Using CONSTA results for US fuel specifications, the pressure vessel equation was converted into a simpler form dependent on temperature and with constants to be recommended for different types of particle configurations. The latter has become part of the GA code SORS (Ref. 27). Pressure vessel failure is not considered significant at temperatures > 1600 °C compared to other failure mechanisms, and is therefore not considered, in this temperature range.

The Russian Federation employs a so-called "ep stress-strain state" model on a more empirical basis to calculate the circumferential stresses in the SiC layer which lead to its failure (Ref. 186). The analytical model equation includes dependencies on stresses upon the SiC layer from inside and outside, deformation, radiation shrinkage and elasticity. Experimental data on shrinkage and elasticity of matrix graphite as function of fast fluence and temperature are available from irradiation tests. Unknown model parameters such as shrinkage and elasticity of the oPyC layer are formulated as polynomials whose constants are then obtained by correlating the equations with the experimental results.

For application of fuel performance models for the lower accident temperature range S 1600 °C, it must be pointed out that the "mathematical" result of a particle failure fraction (which - in PANAMA-I - immediately leads to a gas release fraction) is not necessarily a real failure of a particle. If the calculated value has exceeded a level equivalent to the fraction of one failed particle (which is, for a German spherical fuel element, on the order of 10'4j, it could be interpreted as being composed of contributions from several failed particles having released only

a fraction of their inventories. The full inventory of a particle is only expected to be released at high temperatures.

3.3 RADIONUCLIDE RELEASE FROM COATED PARTICLES UNDER ACCIDENT TEMPERATURE CONDITIONS

3.3.1 Metallic Fission Product Release

Various kinds of modeling approaches are used to describe the release behavior of metallic fission products from coated fuel particles . The main approach used by many authors is based on a diffusion model, and some later models assume a particle coating failure preceeding fission product release. Finally, there are models adopting parts of both modeling approaches.

The calculation model which has been most often used in KFA safety analyses for several types of small and medium-sized HTGRs is the diffusion model FRESCO . This model is available in a core version, code FRESCO-1 (Ref. 29), to determine the fission product release from an HTGR core, and in a pebble version, code FRESCO-11 (Ref. 30), which describes the fission product release from a single spherical fuel element under irradiation (normal operation) and heating (core heatup accident) conditions . Simple changes in the input data allow application to various particle types and fission product species to be considered .

Both FRESCO codes use the same numerical method to determine the radionuclide release from the coated particles in discrete steps of time and lo- cation. Effective$ diffusion coefficients for particle kernel and coating materials are used in the numerical solution of the Fickian diffusion equation. The FRESCO code considers the treatment of two different types of particles : standard particles with an intact coating and exposed particle kernels. The non- releasing character of intact coating layers for gaseous fission products (and

of a fission s "Effective" means that the contributions from all possible transport mechanisms single diffusion coefficient whose product species within a material zone are combined into a A). temperature dependence is given by an Arrhenius relation (see Appendix

ACCIDENTS 47 3.0 HTGR FUEL BEHAVIOR DURING CORE HEATUP iodine) is simulated by a "high" value for the diffusion coefficient in question . A description of collected and recommended data for the diffusion coefficient under accident conditions is given in Appendix A.

A particle failure function in form of a step function is required as input which defines the time points at which a certain fraction of standard particles turn into exposed kernels with the actual fission product inventory in their coatings being immediately released into the fuel element matrix graphite. A particle with a cracked but still existent coating which is observed to better retain (metallic) fission products than an exposed kernel, can be simulated by a lower value for the kernel diffusion coefficient.

Much experience has been gained so far with the KFA diffusion code FRESCO by predicting and postcalculating heating experiments for validation purposes. The codes have been used for core release predictions under core heatup accident conditions for various HTGR designs. The "pebble" version has also been used for predicting the metallic fission product release from coated particles under normal operating conditions. It can be stated that a well assured data set of diffusion coefficients (see Appendix A) is available to make this model an adequate tool for safety analyses (Ref. 12).

In the US fission product release modeling, the "Integrated Failure and Release Model for Standard Particles" (Refs. 28, 163) has replaced the former diffusion model development. This statistical model uses the observed cesium release as indicator for a failure of the SiC layer as mentioned above (see Chapter 3.2), thus providing at the same time a cesium release fraction from the coated particles equal to the failure fraction. Since the model equations have been derived from heating tests with temperatures ~ 1600 °C, the corresponding GA model coded in SORS assumes a certain fission product retention in the particle kernel at temperatures < 1600 °C using a diffusive transport mechanism (Ref. 8) .

The 1988 revision of this joint US/'FRG statistical model has been developed as a replacement of the original version from 1985 as US reference radionuclide release model (Ref. 8). By taking into account the retention capability of matrix graphite even at temperatures between 1600 and 1800 °C as observed in several heating experiments and using a larger experimental data base, the empirical equation for the cesium release from the coated particles is a definite improvement compared to the original version. However, the derived empirical depend-

encies given in the pre-exponential factor for the Weibull distribution do not seem to represent a physical explanation for the observed release behavior.

A further development of the 1985 version of the joint US/FRG statistical model and a companion effort to the on-going revision by Goodin has been made in the Martin-Goodin-Nabielek (MGN) model coded in MACINTOSH (Ref. 31). This model is based on the same SiC failure formalism as in the revised statistical model (SiC thermal decomposition) . But in contrast, the subsequent fission product transport through the failed SiC layer is assumed to proceed as a grain boundary diffusion process within the silicon carbide, with a larger cesium diffusion coefficient required than that recommended in Ref. 1 . The retention effect for cesium in matrix graphite was considered to be important as was done in Goodin's revised version of the statistical model, but was modeled differently. MACINTOSH employs a diffusive transport in the graphite rather than the fast fluence dependent retention factor of the 1988 version.

The MGN model gave good agreement between measured and calculated shapes of cesium fractional release curves from some KFA heating experiments . How- ever, the calculations underpredicted the concentration profiles and, at the same time, overpredicted the release itself.

Two Japanese calculation codes are used to describe fission product release behavior under core heatup accident conditions. Besides the previously mentioned code FORNAX (see section 2.3.1), the code HTCORE has been developed for HTTR safety evaluation purposes (Ref. 34). It is linked to thermohydraulics and nuclear design codes which provide input data for fuel temperature and fission product inventories in the HTTR core. Empirical fission product release rates from the coated particles under HTTR depressurization accident conditions are taken from the GA code SORS as input data. Calculation steps of the HTCORE analysis are the release from the fuel region and diffusive transport through the graphite region into the coolant region . The comparison of model calculations for cesium with isochronal and isothermal heating test data up to 2200 °C have demonstrated a sufficient conservativeness of the code . A modification of the fuel region modeling is planned to allow for a more realistic approach .

The ORNL code SHELL (Ref. 36) is a simplified diffusion model describing the diffusive transport of fission products through a thin spherical shell (SiC layer) for which an analytical solution can be given. A similar approach was used ear-

3.0 HTGR FUEL BEHAVIOR DURING CORE HEATUP ACCIDENTS 49 lier at ORNL described in Ref. 180. Since the model assumes a uniform internal concentration, it is not valid for high fission product release from the particles . The initial conditions at beginning of heating: zero content in coating, no release during irradiation, would be too optimistic (he not conservative). Effective diffusion coefficients derived with this model describe the average transport through kernel and all coating layers at the same time . The SHELL model is usefully applied to cases where there is only one dominant, release rate determining retention barrier, for instance cesium in SiC or krypton in oPyC with a failed SiC layer. For these cases, there is a good agreement between diffusion coefficients derived with SHELL and FRESCO-II; they deviate in cases with additional retention barriers, for instance strontium in both particle kernel and in SiC.

A simple phenomenological model has been developed by the Russian Federation, called activation model (Ref. 189). Fission product release is treated as superposition of releases from two groups of species differing in the activation energy (see also section 2.3.3) .

The recently developed KFA code SPTRAN (Ref. 55) is in its basic aspects a copy of the FRESCO-1 diffusion model. One of the main differences is the method of numerical solution by employing an iteration between fission product concentrations in coated particles and matrix graphite which was later found not to be necessary (Ref. 33). Other differences are the handling of defective particles (one time-dependent "representative" defective particle rather than a step function) and the connection of the core cavern with the primary circuit which, however, is not used for predictions in reactor design studies since it introduces a high degree of uncertainty due to plateout processes (Ref. 33).

Another KFA code, GRECO, (Ref. 3'7) has been developed a couple of years ago which could be regarded as a refined diffusion model comprising a detailed modeling of transport mechanisms in the microstructure of the particle under accident conditions. An independent partial model of GRECO, called GRECO-BREL, calculates fission product release from a spherical fuel element under accident conditions. Compared to simple diffusion codes, these codes need many more input parameters which seem inappropriate to be put into a recommended and validated set of transport data for safety analysis purposes. The GRECO model might be helpful in interpreting postheating examination results.

A collection of actual transport data for various fission product species in HTGR fuel particle materials is given in Appendix A.

3.3.2 Fission Gas and Iodine Release

The release of fission gases and iodine9 from coated fuel particles is widely expected to be directly correlated with a defective or failed coating. This idea im- plies released fission gases (e .g., krypton-85) are the indicator for a broken coating. An intact SiC layer keeps the coated particle gas-tight and an intact oPyC layer causes a significant delay of the transport of gaseous fission products once they have passed the SiC layer.

There are, however, other authors who disagree with the above definition of a defective coating and suggest new ideas for gas transport in silicon carbide (see Myers, krypton transport in SiC, section 2.3.2).

Based on the above "classical" definition of particle failure, the simplest way of determining the gas release fraction is to equate it with the fraction of failed particles assuming a complete and immediate release of the fission gases from the particle upon coating failure. For more refined models, dependencies of fission gas release behavior on different parameters were then incorporated in the modeling of particle failure mechanisms .

A further step in gas release modeling is then to consider a defective SiC layer in combination with an intact oPyC layer, with a diffusive transport of fission gases through the oPyC layer and making cesium the indicator of a SiC layer failure . Cesium is not strongly retained by pyrocarbon at elevated temperatures .

The assumption of the gas release fraction to be equal to the defect fraction is a realistic approximation for high temperatures. However, it does not take account of the retention capability of the fuel kernel at normal operation or lower accident temperature range.

evidence to assume that fission It has been standard practice and supported by experimental accident conditions . gases and iodine show a similar release behavior under

CORE HEATUP ACCIDENTS 51 3.0 HTGR FUEL BEHAVIOR DURING Several heating experiments were dedicated to the study of fission gas release from failed particles. The reduced diffusion coefficients determined from these experiments can be used in diffusion models to calculate the gas release.

A calculation model has been developed in the US by Myers and Morrissey to determine the gaseous fission product fractional release from failed particles under accident conditions as a function of time and temperature (Ref. 73). The analytical function consists of two components, an initial release due to atoms near the exposed surfaces of the kernel escaping without encountering traps, and a slower release of atoms which encounter traps and are delayed in their transport by having to escape from the traps. The diffusion-trapping equation was solved to derive the formal mathematical expressions that are used to describe the release of fission gases, both the initial rapid release as well as the slower release . A good agreement with respect to krypton and xenon isotopes is obtained when comparing with experimental data for BISO Th02 and TRISO HEU UC2 laser- failed particles. Iodine data were found to be reproducible using the xenon parameters. Parameters are also available for cesium, but they require a different interpretation since cesium is released both from the kernel and from the particle coating .

The HRB model of describing the fission gas and iodine release behavior from defective particles during temperature transients is coded in TRAGAS (Ref. 38) . This approach is also based upon the assumption that a simple diffusive transport is not sufficient. Iodine and xenon release data from the German irradiation and heating experiment FRJ2-P28 have been interpreted as a rapid emission from traps in kernel and buffer material, the latter being loaded by recoil from the kernel surface layer . The equation of the fractional release during temperature transients is characterized by two empirical functions, a temperature and time dependent parameter describing the gradual emission from traps and a temperature dependent fraction which escapes the traps in a burst. The burst function due to the recoil release was found to be also dependent on burnup.

The KFA diffusion codes FRESCO-1 and -I1 do not contain a trap model for gases in the particle kernel, but instead use a diffusion coefficient for each species con- sisting of a high temperature and a low temperature branch with different activation energies . This approach conservatively covers the experimental results (see Appendix A).

The JAERI code HTCORE mentioned in the previous section 3.2 .1 is able to predict the noble gas and iodine release from the HTTR core using the corresponding release rate functions from the SORS library.

3.3 .3 Particle Failure Model Discussion

The discussion about the different philosophies upon which the calculation models are based, raises the important question of the definition of a particle failure. In the classical diffusion codes, a defective particle is modeled like an exposed kernel . In the early statistical model, it is modeled like a non-existing SiC layer. Several models try to handle particles with a degraded SiC layer and seem to approximate best the data in many heating experiments. This is possible only on an empirical basis, since the broad range between the least releasing (intact) coating and the most releasing (non-existing) coating must be covered for instance with a factor representing the type of defect (HRB) or with fitted diffusion data. ,

Pro and con arguments could be provided by the IMGA measurements which describe the distribution of fission product inventories of single coated particles in terms of a particle frequency versus inventory fraction diagram (Ref. 76). In theory, a pure diffusion model would create a single line of all particles having retained (or released) the same amount of fission products .1 Q The statistical models also create a single line in the same diagram where a certain fraction of particles has completely released its inventory. Neither model is able to reproduce the existing IMGA results . The assumption of an additional (or hybrid) release mechanism, diffusion through intact particles combined with a particle failure fraction (Ref. 70) or degradation of the SiC layer with a subsequent or enhanced diffusive transport through the degraded silicon carbide (Ref. 71), is one possible explanation. Another is the assumption of an inherent variance of diffusion coefficients to explain the observed variance of remaining cesium fractions in single coated particles (Ref. 74) (see also Appendix A, section A.3) .

would be expected because 10 In reality, some distribution around the average particle inventory distribution of (1) the distribution o£ the SiC layer thicknesses in particles, and (2) an inherent o£ diffusion coefficients due to microstructural variations in the layers.

3.0 HTGR FUEL BEHAVIOR DURING CORE HEATUP ACCIDENTS 53

A calculation model based on effective diffusive transport does not claim to be able to explain all aspects of observed fission product release behavior. However, it appears to be easy to logically combine it with additional processes such as particle failure (diffusive release from exposed kernels), SiC degradation (modeled as a thinning of the SiC layer), fast neutron fluence or fission product concentration dependence of diffusivity.

The advantage of diffusion codes which include the irradiation (normal operation) phase is their ability to commonly describe the results of heating tests with a single diffusion constant if the irradiation history is adequately considered (Ref. 42).

3.4 FISSION PRODUCT TRANSPORT WITHIN THE CORE CAVERN

Fission product retention mechanisms in the fuel element graphite are very important for the release behavior from an HTGR core.

Fuel element graphite serves as a release barrier, at least as a further delay for metallic fission products which either result from heavy metal contamination or have already escaped the fuel particles .

+ Graphite surfaces in colder regions of the core cavern (active core, reflectors) offer sites to re-adsorb metallic fission products which were previously released into the coolant. Adsorption processes are highly efficient at low coolant velocities in reducing the core release during a core heatup accident in a depressurized reactor.

" There is no significant holdup for fission gases and iodine released from defective coated particles nor an adsorption on graphite surfaces at elevated temperatures . However, gaseous fission products and iodine available from heavy metal contamination in the fuel element graphite are believed to be originally sited within the graphite grains. Thus, a slow transport mechanism

is required for them to escape the grain before they are released in a rapid transport via the graphite pores.

There are several models available to describe the radionuclide transport in graphite and its ad-Jdesorption on graphite surfaces .

The two-dimensional KFA diffusion model FRESCO-I (Ref. 29) was developed in order to calculate the fission product release behavior in the core cavern of a pebble-bed HTGR. Simple Fickian diffusion is assumed both in the matrix graphite of the fuel element and in the graphitic reflectors. The separate consideration of diffusion in the graphite grains and in the graphite pores as an uncoupled two-phase diffusion in the matrix graphite is optional.'] In both fuel element and reflector graphite, the consideration of concentration dependent transport data is possible . Concerning the small activity inventory in the graphite grains and in the carbon buffer layer of defective particles, a release by the "slow" diffusive transport to the inner graphite surfaces (grain boundaries) is considered only for lower temperatures c 1250 ° C. For higher temperatures, a spontaneous release of this inventory is assumed due to lack of data (see Appendix A). Main input data for a FRESCO-1 calculation are fuel temperature and coolant gas flow transients, coating failure rates from PANAMA-I, fission product diffusivities and sorption isotherms of fission products on graphitic materials.

While the original version of the Integrated Failure and Release Model for Stand- ard Particles from 1985 completely neglected a retention of cesium in the graphite of the spherical fuel element, the revised version has made the attempt to derive a relation between cesium inventory in the graphite and the fast neutron fluence from German heating experiments. A dependence on fast fluence alone does not seem to be realistic since heating temperature and time are generally expected to play a major role in determining the transport behavior in graphite.

The calculational result of a very low cesium release from the core of small-sized HTGRs due to an almost complete retention by chemisorption in colder graphitic regions remains true, if the experimentally detected deviations from the Fickian

tt The approach of regarding fission product transport in graphite grains and in graphite pores as used in many calculation models is not quite consistent with the fact that most of the fission product inventory is found in the binder component which is not strictly graphite. A diffusion coefficient for the binder material has not been evaluated so far. In this aspect, the diffusive transport in matrix graphite is still unclear.

3.0 HTGR FUEL BEHAVIOR DURING CORE HEATUP ACCIDENTS 55

diffusion model (Refs. 102, 108) are taken into account. Concentration profiles of cesium in graphites with steep gradients near the surface have been found in- dicating that the diffusion mechanism could not be explained by simple Fickian diffusion, but rather required a trapping diffusion model. The trap model does not only consider diffusion sites but also traps within the graphite.

ä cD(x,t) D ö2 C&20cD(2 - * + b (13) - u CD(x,t) -, m(x,t) a r ä x

and

= y " cD(x,t) - 6 " m(x,t) (14)

3 where cD (x,t) : local concentration on diffusion sites [mol/m ] 3 m(x,t) : local concentration on trapping sites [mol/m ] [M2 D : diffusion coefficient /S] k : trapping coefficient [s~ 1] ju _ e /RT Aö -Q'U QK : activation energy [J/mot] b : emission coefficient [s -1] b = b *e-Qb/RT a Qb : activation energy [J/mol]

The trapping rate in graphite is given by:

(15)

where 0t : trapping rate [mol/(m3 s)]

The emission rate from traps is given by:

(16)

where (De: emission rate [mol/(m3 s)]

Values for y and b and the activation energies and Qb have been exper- Qu imentally determined for cesium at low concentrations in A3-3 matrix material (Ref. 102). The diffusion coefficient D of the trap model is correlated to the ef-

fective diffusion coefficient Deff of the Fickian model by:

`_. D Deff (17)

Within this model, the sorption equilibria (represented by the Henry coefficient a) of cesium on graphite are given by:

CPS m + CD

3 where cgas = cesium concentration in gas [moj/m ]

Calculations of sorption kinetics on graphite (initial cesium burden in graphite 0) have been performed with the trapping diffusion model in comparison with the Fickian model using Deff of equation (17). The results indicated that due to the activation effect in trapping, the overall sorbed amount of cesium in graphite at beginning of sorption is smaller in the trapping diffusion model than in the Fickian model by a factor of

57 3.0 HTGR FUEL BEHAVIOR DURING CORE HEATUP ACCIDENTS (19)

However, this difference decreases with increasing cesium burden and goes to zero under equilibrium conditions .

In order to examine the influence of the trapping effect on the overall cesium source term in core heatup events, FRESCO-calculations have been performed with and without the factor 1 /= applied to the experimentally (evaluated from sorption isotherms) determined Henry coefficient a. The results show that the cesium release out of the core cavern in core heatup events of the HTR-Module design is larger by a factor of c 8 if diffusion trapping is taken into account (Ref. 109) . This factor represents an upper limit of the influence of the diffusion trapping for these particular accidents. The overall cesium release out of the core cavern still remains < 10-8 even with this factor and the increase by trapping diffusion remains insignificant from a radiological point of view. Therefore, it may be concluded that the usually applied Fickian diffusion model in graphite is a reasonable tool for safety estimations of small HTGRs. The KFA code FALLDIF (Ref. 153) has been developed to solve the above diffusion equations for cartesian and spherical geometries.

A particularly high sorption capability was found for the ungraphitized binder component in fuel matrix graphite (Ref. 103) . An experimental validation of this retention under accident simulation conditions for gas flow, temperature, and concentration is under way at KFA. The extremely high retention has been calculated only for small-sized HTGRs with their low fission product concentrations in graphite, whereas in medium-sized HTGRs, graphite saturation (Freundlich regime) and chemical reactions like formation of Csl are expected to occur (Ref. 104). Nevertheless, maximum accident temperatures up to 2000 °C might be possible without significant release of cesium and strontium from the core cavern up to 10 d into a core heatup accident, if credit is taken for this chemisorption (Ref. 110) . This maximum temperature corresponds to a thermal power of about 600 MW (Ref. 105). Additional work on this question will be done in the future in order to optimize the concept of the HTR-Module .

With respect to iodine, only a limited potential for source term reduction by chemisorption on cold graphitic components exists in small HTGRs, as was indicated by FRESCO calculations using experimental sorption isotherms for the iodine/graphite system (Ref. 145) . However, the extent of iodine sorption on graphite seems to be strongly influenced by its impurity content (Ref. 106) . Ad- ditional measurements are necessary in order to clarify the applicability of iodine sorption isotherms in HTR accidents.

A detailed description of fission product transport models for reactor graphite has been made in Ref. 41 including a proposition for an extended two-phase diffusion model comprising a concentration-dependent surface diffusion (see Appendix A, section A.4) and a gas phase diffusion which are coupled via sorption isotherms.

Models and data used for sorption on graphitic and metallic surfaces (as used in the FRESCO or SPATRA codes) can also be applied to calculations of temperature induced desorption from graphitic surfaces . Desorption from graphite is only important for the small amount of adsorbed iodine, whereas cesium and strontium are retained by the above outlined chemisorption on cold graphites. Desorption from metal surfaces, in particular from the steam generator, remains small. A core heatup accident does not lead to an overall increase but to an equalization of core temperatures. In most regions of temperature increase, the initial nuclide concentrations are far from saturation which, thus, does not lead to a major desorption.

Desorption from primary circuit surfaces is expected to be insignificant. A reduction of release to the environment occurs due to the buffer effect of the primary circuit volume after depressurization : i.e., the lack of an effective transport mechanism from core to environment. Safety evaluations for German small modular HTGR designs did not consider any depletion along the path into the environment, nor was any retention by the reactor building assumed . Parametric calculations, however, indicate a retention potential for elemental iodine within the reactor building in addition to the above mentioned sorption on graphite. For low gas exchange rates between reactor building and environment after completion of depressurization (fractional exchange rates c 0.1 d" I), a significant source term decrease is expected by depletion of elemental iodine on building surfaces. The same may also hold for silver. Iodine converted to organic com- An pounds like CH3I will not be retained by reactions on building surfaces . 107. overview of depletion rates of iodine compounds on surfaces is found in Ref.

HEATUI' ACCIDENTS 59 3.0 HTGR FUEL BEHAVIOR DURING CORE 60

4.0 HTGR FUEL BEHAVIOR DURING WATER AND AIR 'INGRESS ACCIDENTS

4.1 Fission Product Release from Defective Coated Particles

Attack of oxidants on UO2 in steam and limited air ingress accidents is restricted to particles with defective coatings resulting in a hyperstoichiometric oxide UO 2+X (Refs . 112, 113). During oxidation, a part of the fission product inventory in intergranular bubbles and grain boundaries can be released . In addition, there is an increase of the diffusion coefficients of fission products within the oxide grains. These mechanisms mainly affect the release of iodine and noble gases. There are, however, also some indications that cesium release might be influenced (Ref. 114) .

The equilibrium degree of uranium oxidation is determined by the local oxygen potential which strongly depends on the graphite oxidation processes. Consump- tion of oxidants by graphite decreases the oxygen potential within the fuel element as well as, in steam ingress accidents, the formation of H2 and CO. The first step in release estimations is therefore the calculation of the oxygen potential within the fuel zone of fuel elements during accidents. This is performed with the KFA code COROX (Ref. 115). It calculates graphite oxidation influenced diffusion profiles of the oxidants and of their reaction products in the porous graphite and also estimates the profile of the corrosion zone.

In general, high graphite oxidation rates are coupled with low penetration of oxidants into the graphite and vice versa. Significant rates with penetration depth of c 5 mm are found for the oxygen/graphite reaction at temperatures 650 °C. For the steam/graphite reaction at temperatures > 1050 °C for ambient pressure and at > 1000 °C for normal operation pressures of HTGRs, similar penetration depths are found. These high graphite temperatures are normally required for U02 oxidation for kinetic reasons. Taking into account the fuel-free zone of the spherical fuel elements, a significant fuel oxidation by air is therefore possible only at temperatures below 650 °C. In case of steam, the strong influence of hydrogen on the oxygen potential leads to a further reduction of the upper temperature limit from about 1000 °C down to about 750 °C.

4.0 HTGR FUEL BEHAVIOR DURING WATER AND AIR INGRESS ACCIDENTS 61 With respect to the oxidation kinetics of U02, it is a well known fact that chemical diffusion rates of oxygen in fuel, i.e. diffusion along a gradient of the chemical potential as explained in Ref. 116 are very high. Therefore, a kinetic hindrance of oxidation is not expected as long as the oxidation process on the geometrical fuel surface proceeds sufficiently fast. This seems to be the case for air attack on fuel, but not necessarily for steam . Very slow UO2 oxidation kinetics in steam at 600 °C (Ref. 117) indicate that surface kinetics (or perhaps gas phase kinetics of steam dissociation) are rate limiting steps at low temperatures. In contrast, chemical diffusion controlled rates are found at higher temperatures > 1023 °C (Ref. 118).

In accident analyses, it is assumed that the kinetics of air oxidation of fuel are limited by chemical oxygen diffusion for the whole temperature range. Suffi- ciently fast steam oxidation, however, should take place only at temperatures 600 ° C, again limited by chemical oxygen diffusion. The assumptions for steam ingress are rather conservative.

There is a significant lack of knowledge with respect to the fission product release induced by fuel oxidation, in particular for high burnup (~t 10 % FIMA) fuel . Sufficient knowledge exists only for the increase of in-grain diffusion coefficients of iodine and noble gases due to oxidation, which is two orders of magnitude higher in U02.ta (Ref. 119). In contrast, the fraction of short lived nuclides within intergranular bubbles and on grain boundaries and its release by oxidation is not well known. Heating of high burnup fuel kernels coated only with porous buffer layers up to 1600 °C indicates that the 1-131 inventory outside of U02 grains is c 1 .5 % (Ref. 120). The recoil inventory within the buffer, however, could not be differentiated from the bubble/grain boundary inventory. In accident calculations concerning fuel oxidation, the bubble/grain boundary inventory fraction is conservatively assumed to be 1 .5 % of the overall particle inventory. The inventory in the buffer is neglected here. It is further postulated that the release of the fuel inventory outside grains is proportional to the hyperstoichiometric coefficient x, with nearly complete release at x=0.2. This value has been approximated from release transients of irradiated hyperstoichiometric uranium oxides (Ref, 121) . Additional experimental investigation of the steam attack on irradiated AVR fuel elements with a high (up to 10 %) fraction of defective particles, will be conducted at KFA to improve the knowledge in this area and to reduce conservatism (Ref. 122). Experiments to investigate the oxidation kinetics of U02 kernels in steam/H2 mixtures in the

temperature range 300-700 °C will be performed at FZ Rossendorf. US studies on the hydrolysis of Uß2 fuel are reported in Ref. 161 .

Applying the above outlined model to severe steam ingress accidents of the HTR-Module, the iodine and noble gas release fractions from defective/failed particles and into the environment were found to remain below 3*10"7 of their overall inventories. Release calculations for limited air ingress accidents have not yet been conducted. A significant air ingress into the core of the HTR-Module from a single small leak in the primary circuit enclosure is only possible during a core heatup accident; i.e., when the primary circuit temperatures decrease again and air is sucked in by coolant contraction . Thus, the iodine and gas release into the primary circuit due to fuel oxidation should be much higher for this scenario than for steam ingress accidents. The negligible gas transport from the primary circuit, however, prevents any significant release into the environment.

Most recent experiments conducted in the KORA test facility at KFA (still un- published) have revealed tremendously higher releases than would be estimated by applying the model as described in this section. There is reason to believe that this model may not be as conservative as required for safety analysis purposes.

4.2 Fission Product Release from Graphite

There are two mechanisms for fission product release from oxidized graphite:

" direct interaction of the fission products in graphite with water, steam, or oxygen

" gasification of graphite together with its fission product content

be- Concerning the graphite grain, only the second mechanism is conceivable, cause most of its inventory cannot be reached by the oxidants.

of the graphite, With respect to the inventory on porous inner surfaces graphite oxidation thermochemical equilibrium calculations have indicated that

INGRESS HTGR FUEL BEHAVIOR DURING NVATER AND AIR 4,0 ACCIDENTS 63

by steam is more likely to occur than oxidative cesium or strontium gasification (Ref. 123) . This was confirmed by experiments with steam attack on graphite which contained cesium, strontium, and silver. CO and H2 were detected in the gas phase. The extend of graphite corrosion was deduced to be small.

The same holds true for oxygen attack, as theoretical considerations have shown. The preferential reaction of steam with graphite rather than with chemisorbed metals is not expected at low temperatures (c 750 °C) because this graphite oxidation does not take place for kinetic or thermochemical reasons. However, also under these circumstances, a significant vaporization of cesium sorbed on graphite does not occur, as the equilibrium constant K1 of the following reaction

CS.. .C + H20 -+ CsOHg + 0 .5 H2 ( + C) (20)

on matrix graphite indicates (for thermochemistry of Cs. ..C see Ref. 104):

o PH2 Ki = PCsOH ' o PH20 (21) 44700 + 7.78 CC " e T

where P~i = Pi / Po Pa = 0 .1013 [MPa] s eC ¢ 1 [mmo1C /kgC]

These calculations do not take into account the interaction between gaseous CsOH and steam which is known to occur at high steam pressures. This interaction may lead to an increase of K z , as outlined in detail in Ref. 150. It does, however, not change the result in general . From the thermochemical point of view, a conversion of sorbed cesium into a CsOHS phase is possible only for small H~H20 ratios and high temperatures. For sorbed strontium, a reaction to Sr(OH)2,5 is more favored. Anyway, such oxidations do not increase the release of cesium or strontium from graphite. Another experiment on oxidation induced release from graphite (steam attack) has revealed that the release of chemisorbed strontium remains small compared to the amount of graphite corrosion (Ref. 124). This might be explained by larger migration distances of strontium into the graphite instead of being gasified due to the high stability of its chemisorbed state. This result, however, has conservatively not been taken into consideration in safety analyses up to now.

Attack of liquid water on graphite is expected to mobilize a significant fraction of the chemisorbed metallic fission products. The reason is that ionic metals are much more stable when dissolved in water than as gaseous compounds. Never- theless, the cesium dissolution reaction is endothermic (85 kJ/mol for A3-matrix). Penetration of water into the graphite pores is a slow process as experiments have indicated. In addition, some sorption of Cs + or Sr2+ may occur in presence of liquid water. In safety analyses, chemisorbed metallic fission products are assumed to be completely dissolved from graphite regions penetrated by water.

Concerning the relatively small amount of (weakly) sorbed iodine which is only found at low temperatures with slow graphite oxidation kinetics, a mobilization due to steam, water, or air attack on the graphite must be expected. Low iodine sorption enthalpies account for displacement adsorption to also play a role in this mobilization . A complete mobilization during steam/water attack is assumed for sorbed tritium due to an exchange reaction with water. Tritium, however, is comparably insignificant from a radiological point of view.

The above outlined mobilization kinetics must be considered in context with the fission product distribution in graphite and with the penetration of oxidants/corrosion into the graphite. The KFA code REACT/THERMIX (Ref. 125, see also section 4.3) which also considers chemical reactions between steam and graphite, is used for estimation of the global graphite corrosion and temperature transients. In contrast, local penetration profiles are obtained by the above mentioned COROX code (see section 4 .1). With respect to oxidation of matrix graphite, one should keep in mind that its ungraphitized binder content is preferently oxidized .

core regions Fission products sorbed from the gas phase in the low-temperature graphite, as measure- are concentrated near the outer geometrical surface of the hot core regions, the ments from AVR fuel elements have, indicated (Ref. 126). In is more homoge- distribution of fission products sorbed at the graphtic surfaces

HTGR FUEL BEHAVIOR DURING WATER AND AIR INGRESS 4.0 ACCIDENTS 65 neous. As mentioned in section 3.4, the sorbed fission products are mainly found in the homogeneously distributed binder component of the matrix graphite which amounts to about 10 weight-% of the matrix material. For the inventory of fission products in the graphite grain, there should be a higher concentration in the neighborhood of as-manufactured defective particles caused by uranium diffusion during high-temperature treatment in the fuel element production. In contrast, the activity originating from the natural uranium contamination is equally distributed throughout the matrix graphite. The oxidation induced release fraction of local inventory from reflector graphite is assumed to be equal to its degree of oxidation and from the fuel matrix graphite, to the corrosion degree of its binder content.

Parametric studies for the HTR-Module have indicated that < 0.1 % of the graphite inventory in the core cavern will be gasified in water/steam ingress accidents and an even smaller fraction in (limited) air ingress accidents. Taking that into account, the release fraction from graphite in water or air ingress accidents does not exceed l0" 7 of the overall core inventory of radiological important nuclides.

4.3 Massive Long Term Air Ingress

Long term air ingress with graphite burning (Ref. 146) is possible only by natural convection (chimney draught) from two leaks in top and bottom positions of the pressure vessel and an additional leak in the inner core container; i.e., one large- sized leak in pressure vessel and core container or - in case of the HTR-Module - a complete rupture of the coaxial duct. In addition, there must be a path between the leak(s) in the pressure vessel and the environment. This assumes that the reactor building does not act as a significant barrier against gas exchange. Accident scenarios leading to such configurations are hard to imagine. Their frequency will be < 10-7 Y-1 ; i.e., a very hypothetical regime.

Nevertheless, parametric calculations have been performed for such severe air ingress conditions (Ref. 146) using the code REACT/THERMIX (Ref. 125). Results for severe air ingress in the HTR-Module have predicted the C/4x re-

action front to remain within the bottom reflector whereas fuel element oxidation occurs via the endothermic Boudouard reaction.12 This behavior is typical for accident sequences with a hot bottom reflector at accident initiation. For (the most credible) sequences with gas flow to the core via the (hot) bottom reflector, these calculations indicate a significant time span between start of air ingress and first exposure of coated particles after oxidation of the matrix graphite (which occurs in the active core mainly via the Boudouard reaction). This also holds for high air ingress rates (Ref. 146) restricted only by the flow resistance within the pebble bed core. The exposure of coated particles does not necessarily mean coating destruction and fission product release, because the SiC coating will be converted into Si02 by oxidative attack. Si02 seems to significantly inhibit a further progress of particle corrosion at least at temperatures well below its melting point (> 1600 °C) . For highly corroded regions, the fuel temperatures calculated with REACTJTHERMIX remain below 1500 °C for at least two days. SEM pictures of oxidized (300 h in air at 950 °C) and unoxidized unirradiated coated TRISO particles have revealed that the outer pyrocarbon was removed by oxidation while the SiC layer was not visibly corroded. NUKEM experiments with TRISO particles have shown a small increase of the . fraction of defective SiC layers after heating in air (50 h at 1100 °C) (Ref. 179) . Some irradiated DRAGON TRISO fuel has been burned and no significant increase in the defective coating level has .been detected (5-15 h at 700-1000 °C) (Ref. 148), however, the initial defect fraction was already high (10-2) . The behavior of irradiated fuel elements in air at temperatures up to 1600 °C will be experimentally examined at KFA in order to validate the oxidation resistance of particles under severe air ingress conditions.

The experimental results have shown that despite the worst case assumption that the exposure of particles to oxidation results in the total release of their fission product content, the time span between accident initiation and start of massive the leaks (Ref. release seems .to allow for countermeasures such as covering of coaxial duct 146) . This time span is even larger, if only a complete rupture of the the pressure vessel. is assumed without considering additional leaks at the top of measures should The comparably easy realization of such accident management those accidents. Addi- be taken into consideration in frequency estimations for and graphite tional analytical and experimental work on thermohydraulics

12 Boudouard reaction: C + C02 -Y 2 CO

INGRESS HTGR FUEL BEHAVIOR DURING WATER AND AIR 4.0 ACCIDENTS 67

oxidation in severe air ingress accidents by chimney draught in the HTR-Module are part of future work at KFA.

4.4 Release of Fission Products Plated-Out on Metal Surfaces

4.4.1 Mobilization by Liquid Water

Thermochemical equilibrium considerations indicate that attack of liquid water nearly quantitatively dissolves cesium and strontium which has deposited on metallic/oxidic surfaces (washoff). Cesium plateout data given in Refs. 93, 97 kJ/mol) lead to the following equilibrium for chemisorption in the (AHdes = 234 sub-monolayer regime on lncoloy 800H in presence of liquid water:

(22)Cs. ..M + H20 -* CsOHs,,Iv + 0.5 H2

K2 [ CsOH PHz (23)

[ ] : where molality [mol/kgsolution] a/d : o fractional coverage of geometrical surface with cesium

The cesium washoff reaction is slightly exothermic (-4S kJ/mol) . This is not necessarily true for silver because of its noble character. The sorption of silver on metals has sorption enthalpies similar to the sublimation enthalpy of the element (270 kJ/rnol), which suggests that silver sorption is more like a physi-sorption; i.e., its elemental character probably remains unchanged. Measurements of silver leaching indicate that about 1 % of the deposited amount is dissolved. Never- theless, a silver dissolution of 10 % is conservatively assumed' in recent safety analyses. A total dissolution is postulated for iodine because of its comparably low sorption enthalpy (see section 4.2) and its complex chemical behavior in a liquid solution.

68 Leaching experiments in the temperature range 25-90 °C have indicated that there is a kinetic hindrance of dissolution with time constants of some minutes (Refs . 93, 97, 127) . Similar to the behavior in graphite, this may be caused by the slow transport of water into the pore system of oxide surface layers. This kinetic hindrance is not taken into consideration in safety analyses, because of much higher water temperatures in accidents compared to those in the experiments and because of the poor knowledge with respect to the actual localization of sorbed fission products within the oxide layer and metal.

Besides dissolution and chemical attack of liquid water on sorbed fission products, mechanical (erosive) effects due to a high speed water jet from a tube rupture must also be taken into consideration. The directly impacted surface (some m2) is conservatively assumed to be completely decontaminated .

The fission product content of liquid water is expected to be released mainly by (fast) evaporation of the water (bubbling and droplet formation), followed by volatilization of the residue after complete water evaporation and transport of the gas-borne activity into the environment. In addition, an equilibrium partition of, fission product between steam and liquid water is assumed which is influenced by solubilities in steam (Ref. 150). In the German HTGR safety analyses, it has been conservatively postulated so far that the dissolved fission products are transferred into the gas phase proportional to the degree of water evaporation . There is some evidence, however, that this might be a significant overprediction (Ref. 128), at least for metallic fission products. High gas flow rates with significant water suspension within the gas may have the same effect as evaporation. Data on cesium and iodine washoff are planned to be generated within the next years at FZ Rossendorf as part of the above mentioned research program.

4.4.2 Mobilization by Steam or Air Attack

chemisorbed fission There are few experimental data available on removal of . In addition, the products from metals by an attack of steam (steamoff) or air on HTR relevant data scatter is remarkably high . In steamoff rates for cesium been observed. metals, differences of four orders of magnitude and more have

FUEL BEHAVIOR DURING WATER AND AIR INGRESS 4.0 HTGR ACCIDENTS 69

Thermochemical estimations lead to the following equilibrium relationship for cesium steamoff on Incoloy 800H:

Cs...MIO + H20g --> Cs0Hg + 0.5 H2 (24)

- o P1~Z 3 PCsOH ' ö

_ 16750 _ 4.42 -- 6 T ßo " e

The reaction enthalpy of reaction (24) is endothermic (+ 139 k3/mol). With the ° T W 523 K, and = 0.5 Pa, the assumptions of pH ° = 0.001 *pH o , PH oo equilibrium partial 2pressure of Cs6H is calculated to be on & order of 4* 10-10 interaction is not expected * a/QO Pa. An additional increase by CsOH~steam under these conditions (Ref. 150). In comparison with values of equation (10) (no oxidants present), this is equivalent to an increase of cesium containing species in the gas phase in equilibrium by more than six orders of magnitude due to the presence of steam . If also chemisorption of steam or hydrogen on sorption sites previously occupied by cesium (displacement adsorption) is taken into account, an additional increase is expected which, however, cannot be sufficiently quanti- fied due to lack of data. A still larger increase occurs, if oxidation of metallic sorption sites proceeds simultaneously with the chemical conversion of cesium to CsOH . However, mass transfer in combination with the small partition coefficient between gas and chemisorbed phase significantly limits the relative steamoff rate -r. Under the assumption that sorption is an unactivated process, the ad- /desorption step itself cannot be a rate limiting factor; i .e., a spontaneous ad- /desorption equilibrium on the surface/gas boundary must be expected (Ref. 150) and is considered in the following equation:

Y CsOH z (26) R " T " u

-11 where st relative steamoff rate [s Q : mass transfer coefficient [m/s]

70 should keep One in mind that in equation (26) PCsOHO for the gas bulk is assumed to be zero leading to an upper limit of the mass transfer rate for a given equilibrium partial pressure. Mass transfer coefficients for laminar flow in tubu- lar geometry are on the order of 0.01 m/s which leads from the above given _ 10_ 10-10 I PCsOH ~ 4* 10*a/6o Pa to relative steamoff rates of 1 * s ; i.e., a value just below the lower boundary of that found in Ref. 127. This underestimation of the measured steamoff rate might be caused by uncertainties in experimental desorption enthalpies. A decrease of the desorption enthalpy by 5 % increases the relative steamoff rate by more than one order of magnitude . Uncertainties in mass transfer coefficients and hydrogen partial pressure might also contribute to these deviations . The interpretation given for the steamoff data in Ref. 127, leads to the somewhat surprising result of decreasing steamoff rates with increasing temperatures, and cannot be used in safety analyses without a convincing experimental validation .

The drastically higher steamoff rates of Ref. 129 are partially caused by a larger mass transfer coefficient in these experiments, which may be responsible for an increase of as much as two orders of magnitude. Another rate increasing factor may be a lower sorption enthalpy of cesium on the type of stainless steel as regarded in Ref. 129 due to its composition and surface conditions or due to a fission product burden near its saturation limit. In addition, displacement adsorption or metal oxidation, as outlined above, might play a role in steamoff acceleration . The significant uncertainties in this field require additional experiments. Besides measurements of steamoff rates, the accurate determination of the desorption enthalpies (i.e. , the thermochemicaI stability of the sorbed cesium) is necessary. This is planned for the research program at FZ Rossendorf mentioned above . An algorithm based on equation (26) will be included in the plateout code SPATRA to estimate steamoff rates encountered under accident conditions.

recommended to be In safety analyses, the above outlined calculation method is Considering the uncertainties used for estimating the steamoff rates for cesium . of displacement in chemical behavior, especially the above mentioned influence 1000 should be applied adsorption/adsorbent oxidation, an uncertainty factor of strontium and silver, cesium to the equilibrium partial pressure of CsOH . For . This procedure seems to steamoff rates reduced by a factor of 0.1 could be used . The chemical state of be highly conservative, in particular for strontium

BEHAVIOR DURING WATER AND AIR INGRESS 4.0 HTGR FUEL ACCIDENTS 71

strontium on oxidized metal surfaces is probably not only a chemisorbed one but may be an oxidic phase with volatilities which remain very low even if interaction of its gaseous phase with steam is taken into consideration (Ref. 150) . The lack of experimental data with respect to strontium, however, requires this conservativism .

Cesium mobilization on contact with dry air is small because of the low stability of cesium oxides in comparision with CsOH . In air containing moisture, the reaction

2 Cs.. .MfO + 0.5 Oz + Hz08 -; 2 CsOH (27)

is highly favored (Ref. 150). For strontium and silver in air, a similar procedure as that in steam is applied .

Iodine does not form compounds in reaction with steam or air which are significantly more stable than the element. On the other hand, its chemisorption enthalpies are low and release by displacement adsorption of H2O or oxygen and by adsorbent oxidation might be possible. Evaluation of iodine mobilization in steam/air mixtures indicates that at temperatures > 160 °C a total mobilization can be expected (Ref. 88). Experiments in helium containing a H2O partial pressure of 3000 Pa show that iodine desorption rates at 300 °C increase by about a factor of 20 compared to pure helium (Ref. 130). The dependence of the desorption rate on flow velocity found in these experiments leads to the conclu- sion that mass transfer within the boundary layer is rate determining and not the partition kinetics between gas and solid. Accordingly, the rate increase factors may reflect also the change in equilibrium which will probably be proportional to the steam partial pressure. This effect must be considered in safety analyses, too. Because iodine and steam sorption enthalpies may depend on surface conditions in a different way, a larger increase is expected for other surface conditions than those in the experiments. Therefore, an additional uncertainty factor of 5 is assumed for the above given experimental data leading to the following relation between sorption isotherms in dry and moist helium:

pl,moist° - pl,odry ( 1 . + 3500 " 0 0 ) (28)

The equation (25) is combined with mass transfer correlations in a similar way as shown above for cesium in steam. One should keep in mind that this handling of iodine steamoff contains significant conservatism which should be reduced by additional experiments . An indication of this conservatism may be found in results for the AVR water ingress (Ref. 131). The above mentioned model postu- lates a nearly complete steamoff for iodine in that accident scenario, whereas only about 15 % of the steam generator contamination was found within the water by p° , equation (28) removed from the primary circuit. Substituting p°H O O may be used also for calculating iodine mobilization by air. Reliable data on iodine steamoff at temperatures up to 400 ° C are a goal of the research program at FZ Rossendorf.

4.4.3 Exemplary Source Term Contributions Caused by Mobilization of Plateout Activity

During water ingress accidents in small HTGRs, a mobilization of some percent of the steam generator inventory may be transferred into the gas phase by washoff and subsequent water evaporation. A complete steamoff of iodine is expected for steam attack durations in the order of hours. The steamoff of metallic fission products, however, remains limited to some percent of the steam generator contamination during the depressurization phase. Release of this mobilized activity (fractions of the overall inventory for Cs-137 < 5* 10-6 and for 1-131 < 2*10'7) into the environment is between 10 % and 100 % for relevant scenarios of the HTR-Module. Usually, these water ingress contributions determine the major risk for small HTGRs.

Mobilization by oxygen during a (limited) air ingress accident is at most in the same range as mobilization by steam . The transport into the environment, however, remains very small, because air ingress does not start before completion of primary circuit depressurization, as mentioned above.

4.0 HTGR FUEL BEHAVIOR DURING WATER AND AIR INGRESS ACCIDENTS 73 4.5 Fast Reactivity Transients in Combination with Water Ingress

Although reactivity transients are expected to occur more frequently than the accident scenarios mentioned above, they are still in the hypothetical range (i.e ., low probability) . The temperature behavior in reactivity transients is covered by the above outlined values for core heatup events (Ref. 149) . Therefore, a severe release is not expected to occur in these accidents. However, to simulate some highly hypothetical fast reactivity transients in combination with water ingress, some irradiation experiments were proposed to be performed in the ARGUS pulse reactor (Moscow) on spherical fuel elements. The irradiated fuel spheres will then be heated to 1600 °C at KFA. The data from these experiments are then the basis for the development of a particle failure model for such reactivity accidents.

5.0 HTGR FUEL BEHAVIOR DURING DEPRESSURIZATION ACCIDENTS

5.1 Desorption of Plateout Activity due to Pressure Drop

If an ad-/desorption equilibrium is assumed for all gas/surface interlayers of the primary circuit, a pressure change Ap leads to an increase of the molecular gas- borne equilibrium activity by a factor £p:

Ps ) . Ps f = In ( _ (29) Ps AP AP

where ps : system pressure at depressurization start

Calculations using the KFA code SPATRA, however, indicate that under typical HTGR normal operating conditions, most of the plateout is controlled by mass transfer effects. This means that the surfaces act as a nearly complete sink and that surface concentrations are far from equilibrium, as calculated from the AVR experiment VAMPYR-11 . Nevertheless, due to the comparably small increase in the very low molecular fission product concentration in the cooling gas induced by a pressure drop ( C 5), equation (29) is used in safety analyses to define an p upper limit of the molecular contribution to the depressurization source term.

5.2 Liftoff of 'Dust-Borne Activity

As outlined in section 2.3.4, the uncertainties with respect to dust are remarkably high . In pebble bed HTGRs, most of the dust seems to be produced by friction of fuel element pebbles, in particular in the fuel handling system, and therefore consists mainly of carbon . In addition, some metallic dust has been found containing iron and cobalt and their oxides. Also carburization may play a role in production of some small-sized dust. The AVR reactor contains about 60 kg of

5.0 HTGR FUEL BEHAVIOR DURING DEPRESSURIZATION ACCIDENTS 75

dust after 20 y of operation, as was estimated from results of dust filtering experiments in its hot and cold gas regions. With respect to dust depletion during normal operation, two different effects must be considered :

" dust transport to surfaces mainly by turbulent diffusion and sticking by adhesion

+ sedimentation of (probably larger) dust particles in quiescent coolant regions

It is a well known fact that adhesively bound particles of diameter c 50 hem are not easy to remove by shear forces (Ref. 132). The average size of AVR dust seems to be significantly smaller as indicated by a typical distribution of particle numbers (Refs. 89, 90). The average dust particle diameter of about 0.6 Ym corresponds to a median diameter of about 5-10 ym for a particle volume- weighted distribution . The size distribution of AVR dust seems to be similar for all filter experiments. One should keep in mind, however, that an agglomeration of individual dust particles on surfaces cannot be excluded especially for high dust concentrations, which could produce particle sizes which are more easily lifted-off by depressurization induced shear forces. Adhesively bound dust can change the plateout behavior of primary circuit surfaces .

Sedimented dust is partly lifted-off during fast depressurizations as indicated by AVR experiments with blower transients . Changing the flow rate leads to an increase of the gas-borne dust (which in equilibrium is very low, < 5 Ug/Nm3) by up to three orders of magnitude (Ref. 133). The increase is probably due to flow disturbances during the change of the flow rate. After these disturbances are gone, the increase of dust concentration vanishes by depletion within a few hours. Flow disturbances are also expected during or preceeding depressurization events .

With respect to the dust-borne activity in the AVR, dust from filter experiments was estimated to contain about 10 % of the amount of Cs-137 and Sr-90 released from the core cavern during normal operation (Ref. 91) . Because of the greatly reduced amount of dust expected for modern HTGR concepts such as the HTR-Module, in particular due to dust filtering in the fuel element handling system, the dust-borne fraction of the primary circuit activity is assumed to be smaller by a factor of about 5 for these nuclides as table III (see section 2.3) indicates. Significant activities of Ag-110m were also found in the AVR dust (Ref. 91). In the past, the I-131 content in dust was thought to be nearly negligible. However, recent on-line examination of dust during blower transient experiments

has indicated that the AVR dust mobilized during the transient contains concentrations of 1-131 as large as 3.5 GBq/kg. The total amount of iodine in the AVR primary circuit is on the order of 25 GBq and the total amount of dust is about 60 kg. This means that iodine is inhomogeneously distributed on the dust in the primary circuit and that, in particular, the dust which contains large concentrations of iodine is preferentially mobilized during transients. A possible reason is that the outer layers of the sedimented dust which come into direct contact with the gas flow contain equilibrium fission product concentrations . Considering nuclear decay, there will be a concentration gradient toward the bottom dust layers which settled earliest, especially for short-lived nuclides . The outer dust layers seem to be preferentially mobilized . There is no information available about the activity gradient within the settled dust. Therefore, the dust- borne fraction of the radiologically important iodine nuclides cannot be quanti- fied at this time.

It is important to note that the examination of dust specimens removed by wiping from AVR primary circuit surfaces has shown remarkable differences in the specific activities (Ref. 91) . In particular, dust from fuel element surfaces has a significantly smaller specific activity than dust depleted or adhesively bound on primary circuit components . This was confirmed by the postexamination of irradiated AVR fuel elements, with specific activities on fuel element surfaces (Ref. 126) found to be one to two orders of magnitude smaller than dust activities (Ref. 91). Since dust formation is assumed to proceed mainly by friction of fuel elements, it is not sufficient to consider only the activity on fuel element surfaces as a source of the dust activity.

Interaction of dust with molecular gas-borne or plated-out activity must also be taken into account. This is, however, difficult to correlate with another observation in the AVR: dust from quiescent coolant regions obtained during blower transients has about the same specific activity as adhesively bound dust; i.e., much higher activities than found on fuel element surfaces. A significantly larger fission product loading of dust in quiescent coolant regions is hard to imagine because of the small mass transfer efficiency in this area. This observation can obviously be explained only by the assumption of a (slow) dust particle exchange between areas with sedimented and adhesively bound dust during reactor normal operation. Local flow turbulence (Ref. 134) in combination with the above mentioned growth of adhesively bound particles by agglomeration is possibly re-

7 5.0 HTGR FUEL BEHAVIOR DURING DEPRESSURIZATION ACCIDENTS 7 sponsible for such an activity exchange. These interpretations, however, are highly speculative and cannot be used as a basis for accident calculations.

In AGRs similar experiments have been performed (Ref. 135) . Flow increase by a factor of 2 has been found to raise the level of gas-borne Fe-59, Co-58, and Mn-54 by about four orders of magnitude. These nuclides are assumed to be mainly fixed on dust. In addition, dust particles of 2, 5, and 17 um diameter la- belled with Fe-59 have been injected into the primary circuit . The dependence of depletion and remobilization behavior on dust diameter allows some conclusions about the physical mechanisms. The impact of dust on surfaces seems to be one important depletion mechanism which, however, is reduced by a bouncing-off effect being proportional to particle size and to flow rate. Adhesion sites on the surfaces show a broad distribution of interaction energies with the dust particles due to surface inhomogeneities . This causes remobilization and redepletion of particles until more energetically favorable sites are occupied.

Besides the above outlined plant experience, there are more data available on dust borne activity in gas cooled reactors. Laboratory experiments with graphite dust containing relatively high concentrations of strontium have shown that in the course of dust depletion on hot preoxidized metal surfaces, most of the strontium moves into the metal oxide layer within 1 d, in particular in regions . °C, with high Cr203 content At temperatures c 500 this transfer was less significant (Refs. 136, 137) . Furthermore, these experiments have indicated that dust preferently sticks on clean metallic surfaces, with adhesion forces on oxidized surfaces apparently weaker. A liftoff of the dust particles in the diameter range 50-100 lam by a flow rate increase has been observed . This effect, however, is not representative for HTGR dust because of the comparably large size of the dust used in these experiments . Cesium shows a similar behavior in graphite dust but not as pronounced as strontium.

Liftoff experiments have also been conducted in the French Comedie loop (Ref. 138) . The specimens used in these experiments were taken from the Peach Bot- tom HTGR with block core. Dust in this type of reactor usually shows different characteristics (Ref. 139) than dust from pebble bed HTGRs (i.e ., more metallic and metallic oxide than carbonaceous dust) . Therefore, these results may not be directly applicable . The Comedie experiments indicate a strong dependence of liftoff from surfaces on the shear force ratio. For shear forces smaller than in Peach Bottom normal operation, (almost) no release has been detected . Increas-

ing the shear forces beyond values of Peach Bottom normal operation (shear force ratio = 2) leads to a significant liftoff which is on the order of some percent of the total inventory of cesium, strontium and cobalt on the metallic surfaces. There was no clear distinction between dust and plateout activity in these experiments. Similar tests will be performed in the future as part of the US-MHTGR program in the COMEDIE loop and in the (out-of-pile) DABLE loop at MIT (Boston)6 13

The comparably large liftoff in the COMEDIE experiments raises the question whether a surface erosion with dust formation may occur in the course of HTGR primary circuit depressurizations. Such an erosion could also lead to a partial liftoff of the plated-out activity. With respect to this problem, some cesium plateout and desorption experiments in the KFA loop SMOC should be men tioned. These experiments included a variation of the mass flow (Ref. 140) . In contrast to Comedic, no significant liftoff was observed after flow rate increase. The amount of dust in SMOC is believed to be very small.

Experiments have also been performed to study dust settlement in quiescent coolant regions behind an obstacle (Ref. 141) . Graphitic dust particles with an' average diameter of 50 am have been used. Additional experiments indicate that settled dust is partly lifted-off in case of a reversal of the flow direction.

Summing up the above outlined experience on dust and liftoff, there is no sufficient physical-based assessment available up to now allowing for a reliable description of dust behavior in HTGRs. Some work, however, is in progress in order to obtain a better understanding of the dust behavior.

In actual safety analyses, the adhesively bound dust-borne activity is assumed to be equally distributed over all surfaces of the primary circuit. Since adhesion forces tend to show only a slight temperature dependence, this assumption is not incorrect as long as the surface structure within the primary circuit is more or less homogeneous. The amount of adhesively bound dust is roughly assumed to be half of the total dust amount. The other half is considered to be sedimented dust . AVR data are the basis for the dust-borne fractions of the overall activity given in table III (see section 2.3) . For long-lived nuclides, these values consider a smaller relative dust production rate in the HTR-Module compared to the AVR

13 According to a recent infonnation, the DABLE loop will not be used.

5.0 HTGR FUEL BEHAVIOR DURING DEPRESSURIZATION ACCIDENTS 79 due to a dust filter in the refueling system. Sufficient AVR data for iodine do not exist. The relative iodine sorption capacity of dust is roughly assumed to be 50 % of that of the long-lived metals. This value considers both the weaker iodine sorption on graphite and its stronger sorption on metallic dust. A reduction of the dust-borne fraction of iodine due to the smaller dust formation rates in modern HTGRs compared to the AVR is not taken into account for short-lived nuclides because their interaction with dust seems to be restricted to the dust layers in contact with the coolant.

In case of a depressurization, the dust in areas with shear forces larger than in normal operation conditions is predicted to be completely lifted-off. With respect to the release of dust settled in quiescent coolant regions by sedimentation, 1 % of this dust inventory is additionally assumed to be lifted-off in case of depressurizations where flow disturbances in these regions cannot explicitly be excluded . This value is at least a factor of 5 larger than observed in experiments with blower transients (AVR, CAGR), but has been chosen because of the inad- equate knowledge about the liftoff effect of settled dust. The concentrations of long-lived nuclides are assumed to be equally distributed in the dust settled by sedimentation, whereas short-lived iodine isotopes are conservatively assumed to be completely sorbed in the mobilized dust fraction. Part of the adsorbable activity entering the gas purification system is sorbed on dust particles collected on filters. In safety analyses, 20 % of all adsorbable activities reaching the purification system is assumed to be dust-borne at the beginning of a depressurization accident. These inventories, whose absolute values also depend on the cleaning capacity of the particular purification system, are significant in depressurizations with rapid, reverse flow in the purification system (depressurization via a leak).

5.3 Source Terms in Depressurization Events

The amount of I-131 which may be released during a (dry) primary circuit depressurization of the HTR-Module without dust mobilization is on the order of 0.02 GBq. In case of accidents with mobilization of sedimented dust in the primary circuit and complete blow-out of the dust inventory of the gas purification system, recent work indicates the 1-131 and Cs-137 release into the environment is conservatively estimated to be 2 GBq and 0.4 GBq, respectively. The

contribution from dust in the purification system dominates the Cs-137 source term (about 80 %), but remains negligible for iodine . A significant release of adhesively bound dust is not expected for depressurizations with leak diameters in the primary circuit enclosure S 0.065 m (refueling tube diameter), because in most primary circuit areas, shear forces are below that of normal operation conditions. Larger leaks are much less frequent due to design measures. The contribution of these small source terms to the overall risk of the HTR-Module is significant because of the relatively high frequency of these events.

5.0 HTGR FUEL BEHAVIOR DURING DEPRESSURIZATION ACCIDENTS 8 1 82

6.0 ACKNOWLEDGEMENT

The FRG portion of this work has been performed with funding from the German Federal Government and from the State Government of Northrhine Westphalia. The US contribution has been sponsored by the Office of Advanced Reactor Programs, Division of HTGRs, U.S. Department of Energy, under con- tract DE-AC05-84OR21400 with Martin Marietta Energy Systems, Inc.

The authors gratefully acknowledge the thorough review of this report by B. F. Myers.

6.0 ACKNOWLEDGEMENT 83 84

7.0 REFERENCES 1 . R. MOORMANN and K. VERFONDERN, "Methodik umfassender probabilistischer Sicherheitsanalysen für zukünftige HTR-Anlagenkonzepte - Ein Statusbericht (Stand 1986), Band 3: Spaltproduktfreisetzung", Jül-Spez-388/Vol. 3, Research Center Jülich (1987). 2. E. BALTHESEN, "HTR Fuel Element Development in the Federal Republic of Germany", Proc. IAEA Specialists' Meeting on Behaviour of GCR Fuel under Acci- dent Conditions, held at Oak Ridge TN, Nov. 5-8, 1990, IAEA, IWGGCR/25, Vienna (1991), pp . 24-25. 3. H. NABIELEK, K. FUKUDA, and M. J . KANIA, "HTR Fuels for Gas-Turbine Applications", Proc. GCRA Int. Workshop on the Closed-Cycle Gas-Turbine Modular High-Temperature Gas-Cooled Reactor, held at Boston MS, June 17-19, 1991, MIT, Cambridge (1991), pp. 4-8 - 4-14. 4. T. D. DUNN, K. DATTA, and C. HOOT, "Preliminary Safety Assessment of the 450 MWT MHTGR", Proc. IAEA Specialists' Meeting on Behaviour of GCR Fuel under Acci- dent Conditions, held at Oak Ridge TN, Nov. 5-8, 1990, IAEA, IWGGCR/25, Vienna (1991), pp. 111-I 16. 5. S. SHIOZAWA, "Overview of Current Research and Development Programmes for Fuel in Japan", Proc. IAEA Specialists' Meeting on Behaviour of GCR Fuel under Acci- dent Conditions, held at Oak Ridge TN, Nov. 5-8, 1990, IAEA, IWGGCR/25, Vienna (1991), pp . 19-21. 6. V. N. GREBENNIK, "Review of Current Work on Investigation of HTGR Fuel in the Union of Soviet Socialist Republics", Proc. IAEA Specialists' Meeting on Behaviour of GCR Fuel under Acci- dent Conditions, held at Oak Ridge TN, Nov. 5-8, 1990, IAEA, IWGGCR/25, Vienna (1991), pp. 25-27. 7. XU SHIJIANG et. al., "Fission Product Release Experiment of Coated Fuel Particles", Private Communication, July 1991 . 8. K. VERFONDERN, T. D. DUNN, and J . M. BOLIN, "Comparison of US/FRG Accident Condition Models for HTGR Fuel Failure and Radionuclide Release", J01-2458, Research Center Jülich (1991). 9. G. W. MOMOT and KHRULJEV, "Limits of Safe Operation in Terms of Fission Product Release from the Fuel Element and its Justification",

7.0 REFERENCES 85

Proc. ist Soviet/German Seminar on Fuel and Graphite for HTR, held at Moscow, Russia, July 23-27, 1990, KFA Internal Report KFA-HTA-IB-6/90, Research Center Jülich (x99(3), pp. 319-335. 10. W. KOHNLEIN, "Fission Product Release under Normal Operating Conditions", IAEA Technical Workshop on Unification of Coated Particle Performance Models and Fission Product Transport Data for the HTR, held at Jülich, FRG, Dec. 2-4,1991 . 11. S. SAITO et. al., "Safety Requirements and Research and Development on HTTR Fuel in Japan", Proc. IAEA Specialists' Meeting on Behaviour of GCR Fuel under Acci- dent Conditions, held at Oak Ridge TN, Nov. 5-8, 1990, IAEA, IWGGCR/25, Vienna (1991), pp . 31-36. 12. K. VERFONDERN, W. SCHENK, and H. NABIELEK, "Passive Safety Characteristics of Fuel for a Modular High-Temperature Reactor and Fuel Performance Modeling under Accident Conditions", Nucl. Techn. 91 (1990), pp. 235-246. 13. H. RAGOSS, "Kontamination und Partikelbruch im Modul-Referenz-Brennelement", Interatom Technical Note No . 76.00733.1, Interatom (1989). 14. S. A. BALANKIN and E. A. RYBAKOVA, "Mathematical Model of the Behavior of Oxide Microfuel-Elements", Atomnaya tnergiya 58 (1985), No. 2, pp. 101-104. 15 . A. H. BOOTH, "A Method of Calculating Fission Gas Diffusion from U02 Fuel and its Application to the X-2-F Loop Test", CRDC-721, Atomic Energy of Canada Limited (1957) . 16. G. H. LOHNERT, H. NABIELEK, and W. SCHENK, "The Fuel Element of the HTR-Module, a Prerequisite of an Inherently Safe Reactor", Nucl. Eng. Des. 109 (1988), pp. 257-263. 17. U. H. MATTKE, "Berechnung der Spaltproduktverteilung in Care und Primärkreislauf eines Hochtemperaturreaktors im Normalbetrieb", Ril-2515, Research Center Jülich (1991) . 18. A. MOLLER, "STADIF-II, ein Programm zur Berechnung der stationären Kühlgasaktivität in Kugelhaufen-HTR", Jül-1464, Research Center Jülich (1977) . 19. D. T. GOODIN, "A Single Model for the Performance of HEU/LEU Fissile and Th02 Fer- tile Fuel under Hypothetical Accident Conditions", GA-A16291, General Atomic Company (1981) .

20. D. T. GOODIN, "Accident Condition Performance of High-Temperature Gas-Cooled Reac- tor Fuels", GA-A16508, GA Technologies (1983) . 21 . K. VERFONDERN and H. NABIELEK, "PANAMA - Ein Rechenprogramm zur Vorhersage des Partikelbruch- anteils von TRISO-Partikeln unter Störfallbedingungen", Jül-Spez-298, Research Center Jülich (1985). 22. F. C. MONTGOMERY, "Fission Product SiC Reaction in HTGR Fuel," GA Document No. 905837, General Atomic Company (1981) . 23. E. WALLURA et. al., "Siliciumcarbid-Hüllschichten von Kernbrennstoffteilchen - eine Charak- terisierungsstudie", Jül-1871, Research Center Jülich (1983) . 24. R. BENZ, "Kinetics of Decomposition of CVD SiC in Modified TRISO-Coated Fuel Particles at Temperatures of 1600 to 2000 °C", Technical Note KFA-IRW-TN-124/82, Research Center Jülich (1982) . 25. T. OGAWA et. al., "A Model to Predict the Ultimate Failure of Coated Fuel Particles during Core Heatup Events", Nucl. Techn. 96 (1991), pp. 314-322. 26. W. J. KOVACS, K. BONGARTZ, and D. T. GOODIN, "TRISO Coated HTGR Fuel Pressure Vessel Performance Models", GA-A16807, GA Technologies (1983) . 27. M. H. SCHWARTZ, D. B. SEDGLEY, and M. M. MENDONCA, "SORS: Computer Programs for Analyzing Fission Product Release from HTGR Cores during Transient Temperature Excursions," GA-A12462, General Atomic Company (1974) . 28. H. NABIELEK, K. VERFONDERN, and D. T. GOODIN, "HTR Fuel: Prediction of Fission Product Release in Accidents", Proc. IAEA Specialists Meeting on Fission Product Release and Transport in Gas-Cooled Reactors, held at Berkeley/UK, Oct. 22-25, 1985, IAEA IWGGCR/13, Vienna (1986), pp . 155-169 . 29. H. KROHN, "Freisetzung von Spaltprodukten aus dem Core eines Kugelhaufenreaktors bei Störfällen mit Core-Aufheizung", Jü1-1791, Research Center Jülich (1982) . 30. KROHN and R. FINKEN, H. Berechnung der "FRESCO-II: Ein Rechenprogramm zur Spaltproduktfreisetzung aus kugelförmigen HTR-Brennelementen in Bestrahlungs- und Ausheizexperimenten", Jül-Spez-212, Research Center Jülich (1983) .

7.0 REFERENCES 87

31. J. L. MARTIN and H. NABIELEK, "Modelling of Fuel Element Heating Tests", Technical Note KFA-HBK-TN-4/88, Research Center Jülich (1988). 32. B. F. MYERS, "Selective Aspects of Transport in SiC", Proc. US/FRG Experts' Meeting on MHTGR/HTR Fuel Performance under Accident Conditions, held at Oak Ridge TN, June 8-9, 1989, KFA Internal Report KFA-HTA-IB-2/90, Research Center Jülich (1990), pp. 199-222. 33. L. KRANZ, "Spaltprodukttransport im . Hochtemperaturreaktor : Theorie, programm- technische Umsetzung und berprüfung an Experimenten", Jül-2463, Research Center Jülich (1991). 34. K. SAWA et. al., "Analytical Method and Result of Fission Product Release from Core During a Depressurization Accident of HTTR", Proc. IAEA Specialists' Meeting on Behaviour of GCR Fuel under Acci- dent Conditions, held at Oak Ridge TN, Nov. 5-8, 1990, IAEA, IWGGCR/25, Vienna (1991), pp. 117-123. 35. K. SAWA et. al., "Analytical Method of Fractional Release of Fission Products from Fuel Elements of HTTR", Proc. IAEA Specialists' Meeting on Behaviour of GCR Fuel under Acci- dent Conditions, held at Oak Ridge TN, Nov. 5-8, 1990, IAEA, IWGGCR/25, Vienna (1991), pp. 55-61 . 36. R. C. MARTIN, "Diffusion Modeling of Fission Product Release during Depressurized Core Conduction Cooldown Conditions", Proc. IAEA Specialists' Meeting on Behaviour of GCR Fuel under Acci- dent Conditions, held at Oak Ridge TN, Nov. 5-8, 1990, IAEA, IWGGCR/25, Vienna (1991), pp. 77-84. 37. T. A. BATALAS, N. INIOTAKIS, and C. B. VON DER DECKEN, "Theoretische Untersuchungen zur Spaltproduktfreisetzung aus dem Core eines Hochtemperaturreaktors bei hypothetischen Coreaufhei- zungsstörfällen am Beispiel von Cäsium", Jül-2045, Research Center Jülich (1986) . 38. A. CHRIST, "Transiente Jodfreisetzung aus UO -Defektpartikeln", BBC-HRB Technical Note I-K-5125-BF-GHRA 001785, HRB Mannheim (1986) . 39. D. G. HURST, "Diffusion of Fission Gas, Calculated Diffusion from a Sphere Taking into Account Trapping and Return from Traps", AECL-1550, Atomic Energy of Canada Limited (1962) .

40. T. OGAWA et. al ., "Release of Short-Lived Noble Gases from HTGR Fuel with Failed Coated Fuel Particles and Contaminated Matrix", Nucl. Eng . Des . 132 (1991), pp. 31-37. 41 . V. DANNERT, "Mehrphasendiffusion von Spaltprodukten in Graphit - Ein Beitrag über die Sicherheit von Hochtemperaturreaktoren", Jül-2277, Research Center Jülich (1989). 42. A. CHRIST, "Ausheizexperimente oberhalb 1600 "C", BBC-H.RB Technical Note HTR-5125-BF-GHRA 005058, HRB Mannheim (1987). 43. D. T. GOODIN, "Metallic Fission Product Release Models for Accident Conditions", GA Document No. 906957, GA Technologies (1983) . 44. W. SCHENK and H. NABIELEK, "Kugelbrennelemente mit TRISO-Partikeln bei St8rfalltemperaturen", Jül-Spez-487, Research Center Jülich (1989). 45. V. DANNERT, "Ein thermodynamischer Ansatz zur Spaltproduktrückhaltung - Entlastung von Barrieren", Proc. Jahrestagung Kerntechnik '91, held at Bonn, May. 14-16, 1991, In- forum GmbH, Bonn (1991), pp. 107-110. 46. A. N. GUDKOV et. al., "Behavior of Solid Fission Products in Coated Fuel Particles of a High- Temperature Gas-Cooled Reactor", Atomnaya Energiya 67 (1989), No . 2, pp. 93-97. 47. P. R. KASTEN, M. R. CORUM, and P. L. RITTENHOUSE, "Research on Very High Temperature Gas Reactors", EPRI ER/NP-7372, Electric Power Research Institute, Palo Alto (1991). 48. A. CHRIST, "Nachrechnung von Ausheizexperimenten", BBC-HRB Technical Note HBK-5125-BF-GHRA 001555, HRB Mannheim (1985) . 49. W. HEIT, "Matrixkontamination der THTR-Brennelemente", Private Communication, October 1987. 50. M. J. BELL, "ORIGEN - The ORNL Isotope Generation and Depletion Code", ORNL-4628, Oak Ridge National Laboratory (1973) . 51 . K. VERFONDERN, H. NABIELEK, and W. SCHENK, "Partikelbruch und Spaltproduktfreisetzung in Brennelementen mit (Th,U)O HTI BISO Partikeln bei Betriebs- und St6rfallbedingungen", Internal ieport KFA-ISF-IB-4/88, Research Center Jülich (1988) .

7.0 REFERENCES 89

52. F. C. MONTGOMERY, K. E. PARTAIN, and H. C. ALLEN, "Kernel Migration in Failed, Oxide-Based HTGR Fuel", GA-A17559, GA Technologies (1984) . 53. O. M. STANSFIELD, W. A. SIMON, and A. M. BAXTER, "Fuel Performance Models for High-Temperature Gas-Cooled Reactor Core Design", GA-A16982, GA Technologies (1983) . 54. KFA-HBK, "Projektbericht 1985", Internal Report KFA-HBK-IB-1/86, Research Center Jülich (1986). 55. R. BERGER-ROSSA, "Rückhaltung und Umverteilung von Spaltprodukten im Core und im Primärkreislauf von Hochtemperaturreaktoren bei hypothetischen Störfallereignissen", Jül-2188, Research Center Jülich (1988) . 56. J. THIEL, "Gasfreisetzung aus Defektpartikeln im Experiment FRJ2-P25", HRB Report 232-520-DBI583, HRB Mannheim (1982). 57. H . NABIELEK and B. F. MYERS, "Fission Product Retention in HTR Fuel", Proc. BNES Conference on Gas-Cooled Reactors Today, held at Bristol/UK, Sept. 20-24, 1982, British Nuclear Energy Society, London (1982), pp. 145-152. 58. J. A. TURNBULL et. al., "The Diffusion Coefficients of Gaseous and Volatile Species During the Irradiation of Uranium Dioxide", J. Nucl. Mat. 107 (1982), pp. 168-184. 59. K. RÖLLIG, "The THTR Coolant Gas Activity, an Indicator of Fuel Performance", Proc. IAEA Specialists' Meeting on Behaviour of GCR Fuel under Acci- dent Conditions, held at Oak Ridge TN, Nov, 5-8, 1990, IAEA, IWGGCR/25, Vienna (1991), pp. 99-108. 60. U. WAWRZIK and G. IVENS, "Operational Monitoring of the Release Behavior of the AVR Core", Proc. IAEA Specialists Meeting on Fission Product Release and Transport in Gas-Cooled Reactors, held at Berkeley/UK, Oct. 22-25, 1985, IAEA IWGGCR/13, Vienna (1986), pp . 15-26. 61 . D. L. HANSON, V. JOVANOVIC, and R. D. BURNETTE, "Comparison of Predicted and Measured Fission Product Behavior in the Fort St. Vrain HTGR During the First Three Cycles of Operation", Proc. IAEA Specialists Meeting on Fission Product Release and Transport in Gas-Cooled Reactors, held at BerkeleyjUK, Oct, 22-25, 1985, IAEA IWGGCR/13, Vienna (1986), pp . 2-14.

62. W. W. HUDRITSCH, V. JOVANOVIC, and D. L. GEORGHIOU, "SURVEY, a Computer Code for the Thermal and Fuel Performance Analysis of High-Temperature Gas-Cooled Reactors, User's Manual", GA-C17554, GA Technologies (1984) . 63. K. FUKUDA, T. SAWAI, and K. IKAWA, "Diffusion of Fission Products in Matrix Graphite for VHTR Fuel Com- pacts", J. Nucl. Sci. Tech. 21 (1984), pp. 126-132. 64. K. HAYASHI et. al., "In-Pile Release Behavior of Metallic Fission Products in Graphite Materi- als of an HTGR Fuel Assembly", J. Nucl. Mat. 149 (1987), pp . 57-68 . 65. P. D. SMITH, "TRAFIC, A Computer Program for Calculating the Release of Metallic Fission Products from an HTGR Core", GA-A14721, General Atomic Company (1978) . 66. A. O. GOL'TSEV and A. A. KHRULEV, "A Computational Study of Factors Affecting HTGR Primary Circuit Ac- tivity", Proc . JAIF-GKAE Seminar on Fuel Elements and Fuel Composition of HTGR, held at Tokyo/Japan, Oct. 20-22, 1987, JAIF (1987), pp. 2-50 - 2-68. 67. A. S. CHERNIKOV et. al., "HTGR Coated Particles, Coating Material Properties and Prereactor Test Results", Proc. JAIF-GKAE Seminar on Fuel Elements and Fuel Composition of HTGR, held at Tokyo/Japan, Oct. 20-22,1987, JAIF (1987), pp. 3-1 - 3-20. 68. M. TOKAR, "Evaluation of High Temperature Gas Cooled Reactor Fuel Particle Coat- ing Failure Models and Data", US Nuclear Regulatory Commission Report NUREG-0111 (1976). 69. C. L. SMITH, "Fuel Particle Behavior under Normal and Transient Conditions", GA-A12971, General Atomic Company (1974). 70. K. VERFONDERN, "Possible Explanation for HFR-K3/3 IMGA Results", ORNL/M-2248, Oak Ridge National Laboratory (1991). 71 . R. C. MARTIN, "Analysis of Fission Product Release Data for German Fuel Sphere HFR-K3/3", ORNL Draft Report.

7.0 REFERENCES 9 1 72. 0. BABA, "Evaluation of the Fuel Failure Fraction in HTTR", IAEA Technical Workshop on Unification of Coated Particle Performance Models and Fission Product Transport Data for the HTR, held at Jülich, FRG, Dec. 2-4,1991 . 73. B. F. MYERS and R. E. MORRISSEY, "Licensing Topical Report: The Measurement and Modelling of Time- Dependent Fission Product Release from Failed HTGR Fuel Particles under Accident Conditions", GA-A15439, General Atomic Company (1980). 74. K. RÖLLIG, "Capabilities of the Diffusion Model for Metallic Fission Products at Tem- perature Transients up to 1600 °C - Case: Cs-137 Release from Spherical Fuel Element R2-K.13/1", IAEA Technical Workshop on Unification of Coated Particle Performance Models and Fission Product Transport Data for the HTR, held at Jülich, FRG, Dec. 2-4, 1991 . 75. K. VERFONDERN and D. MÜLLER, "FRESCO-II, Programm zur Berechnung der Spaltproduktfreisetzung aus einem HTR-Brennelement im Normalbetrieb (Bestrahlungsexperiment) und Kernaufheizstörfall (Heizexperiment) - Benutzerhandbuch", Internal Report KFA-ISR-1B-4j91, Research Center Jülich (1991). 76. C. A. BALDWIN and M. J. KANIA, "Fission Product Retention in TRISO Coated U02 Particle Fuels Subjected to HTR Simulated Core Heating Tests", Proc. IAEA Specialists' Meeting on Behaviour of GCR Fuel under Acci- dent Conditions, held at Oak Ridge TN, Nov. 5-8, 1990, IAEA, IWGGCRf25, Vienna (1991), pp. 132-138. 77. A. MÜLLER, "Freisetzung gasf6rmiger Spaltprodukte (Kr, Xe, J) aus Brennelementen für gasgekühlte Hochtemperdturreaktoren", Jal-1295, Research Center Jülich (1976). 78. INSTITUTE FOR NUCLEAR ENERGY TECHNOLOGY (INET), "Progress Report on HTR R&D Program in 1990", Institute Nuclear Energy Technology (INET), PR China (1991) . 79. W. REHM et . al., "Analysen zum Temperatur- und Spaltproduktverhalten des THTR-300 bei Coreaufheizunfällen", in "Untersuchungen zu ausgewählten risikobestim- menden Ereignisabläufen für den Thorium-Hochtemperatur-Reaktor THTR-300 in Hamm-Uentrop", GRS-A-1412, Gesellschaft für Reaktorsicherheit, Köln (1988) . 80. W. JAHN, M. HABEN, and W. REHM, "Corethermodynamik", in "Sicherheitstechnische Untersuchungen zum St6rfallverhalten des HTR-500", Jül-Spez-240, Research Center Jülich (1984) .

81 . SIEMENS/INTERATOM, "Aktivierungsfreisetzung und Strahlenexposition bei der HTR-Modul- Kraftwerksanlage, Teil II : Störfälle," Technical Report No. 76 .00561 .4, Interatom (1987) . 82. W. SCHENK and H. NABIELEK, "Iodine Retention in Spherical Fuel Elements under HTR MODUL Acci- dent Conditions", Proc. Jahrestagung Kerntechnik '90, held at Nürnberg, May. 15-17, 1990, Inforum GmbH, Bonn (1991), pp. 363-366. 83. F. E. VANSLAGER, "RAD 2, A Computer Program for Calculating Fission Product Radio- activities", GAMD-6519, General Dynamics, Gulf General Atomic (1965). 84. R. D. BURNETTE, "Analysis of the First Plateout Probe from the Fort St. Vrain HTGR", Trans. Am. Nucl. Soc. 46 (1984), pp. 379-380. 85. R. MOORMANN and G. MEISTER, "SpaItproduktablagerung auf Metallen und der Code SPATRA", Proc. Seminar on Spaltproduktverhalten in Kernreaktoren, held at Rossendorf/FRG, June 4-6, 1991, ZFK-747, FZ Rossendorf (1991), pp. 56-60. 86. N INIOTAKIS, "Rechenprogramm PATRAS; Programmbeschreibung", Internal Report KFA-IRB-IB-7/84, Research Center Jülich (1984). 87. T. S. KRESS and F. H. NEILL, "A Model for Fission Product Transport and Deposition under Isothermal Conditions", ORNL-TM-1274, Oak Ridge National Laboratory (1965) . 88. E. HOINKIS, "A Review of the Adsorption of Iodine on Metal and its Behaviour in Loops", ORNL-TM-2916, Oak Ridge National Laboratory (1970) . 89. P. BIEDERMANN and B. SACKMANN, "Labor- und Reaktorversuche zur Ablagerung von Spaltprodukten im HTR-Primärkreislauf", Jül-Conf-64, Research Center Jülich (1988), pp. 72-87. 90 . D. L. HANSON, N. L. BALDWIN, and D. E. STRONG, "Fission Product Behaviour in the Peach Bottom and Fort St. Vrain HTGR's", Proc. IAEA Specialists Meeting on Coolant Chemistry, Plate-out and De- contamination in Gas-Cooled Reactors, held at Jiilich/FRG, Dec. 2-4, 1980, IAEA IWGGCR/2, Vienna (1981), pp. 49-54 91 . U. WAWRZIK and C. B. VON DER DECKEN, "Staub- und Aktivitätsverhalten", VDI-Berichte 729 (1989), pp. 239-254.

7.0 REFERENCES 93 92. F. E. VANSLAGER and L. D. MEARS, "PAD: A Computer Code for Calculating the Plateout Activity Distribution in a Reactor Circuit", GA Document No. 10460, Gulf General Atomic (1971) . 93. A. CHRIST, F. W. HAASE, and J . WIEDMANN, "Experimentelle Untersuchungen zur Cäsium-Ablagerung auf Hochtemperatur-Werkstoffen", Proc. Jahrestagung Kerntechnik '85, held at München, May, 21-23, 1985, Inforum GmbH, Bonn (1985), pp. 175-178. 94. H. J . MATZKE and S. PICKERING, "Diffusion of Cesium in Stainless Steel and Possible Implications for Cs Depletion and Mobility", J. Nucl. Sci. Techn. 20 (1983), pp. 237-245. 95. K. HAMMEKE, "Experimentelle Ergebnisse der Ablagerungs- und Diffusionsmessungen von Spalt- und Aktivierungsprodukten beim Bestrahlungsexperiment Saphir 11 am Reaktor PEGASE in Cadarache", Internal Report KFA-IRB-IB-22/83, Research Center Jülich (1983) . 96. 1. PETERSEN, "Zusammenstellung aller experimentellen Ergebnisse der Plate-out Untersuchungen an dem Gasführungsrohr des KFA-Heißgas- isolierungsexperiments im DRAGON-Reaktor", Internal Report KFA-IRB-1B-12/84, Research Center Jülich (1984). 97. K. RÖLLIG et. al., "Cesium Deposition on HTR Primary Circuit Materials", Proc. IAEA Specialists Meeting on. Fission Product Release and Transport in Gas-Cooled Reactors, held at Berkeley/UK, Oct. 22-25, 1985, IAEA IWGGCR/13, Vienna (1986), pp. 329-350. 98. K. D. EHRHARDT, "PLATO, ein Programm zur Berechnung der Spaltproduktablagerung", Jül-Spez-135, Research Center Jülich (1981) . 99. K. P. STEWART, "Final Summary Report on the PEACH BOTTOM End of Life Program", GA-A14404, General Atomic Company (1978). 100. D. L. HANSON, "Results of the General Atomic Deposition Loop Program", GA-A13140, General Atomic Company (1976). 101 . D. RETTIG and H. STÖBEL, "Arbeitsprogramm zum Spaltproduktverhalten", Proc . Seminar on Spaltproduktverhalten in Kernreaktoren, held at Rosendorf/FRG, June 4-6, 1991, ZFK-747, FZ Rossendorf (1991), pp. 61-66. 102. H. GAUS et. al., "Untersuchungen zum Cs-Transport in Graphit", HMI-B 315, Hahn-Meitner-Institut Berlin (1979).

103. K. HILPERT et. al., "Sorption of Cesium and its Vaporization from Graphitic Materials at High Temperatures", High Temperatures-High Pressures 20 (1988), pp. 157-164. 104. R. MOORMANN and K. HILPERT, "Chemical Behavior of Fission Products in Core Heat-up Accidents of High-Temperature Gas-Cooled Reactors", Nucl. Techn. 94 (1991), pp. 56-67. 105. U. FRICKE, "Untersuchungen zur Leistungssteigerung von inhärent sicheren HTR durch Optimierung der Coreauslegung", Jül-2155, Research Center Jülich (1987) . 106. A. SCHOOFS, "Die physikalischen Grundlagen der Absorption und Desorption von Jod in Reaktorgraphit", Diploma Thesis, RWTH Aachen (1989). 107. E. C. BEAHM, W. E. SHOCKLEY, and C. F. WEBER, "Chemistry and Transport of Iodine in Containment", Proc . Int. Symposium on Source Term Evaluation for Accident Conditions, held at Columbus, OH, Oct. 28-Nov. 1, 1985, pp . 479-500. 108. D. W. HILL and B. F. MYERS, "Temperature, Pressure and Oxidation Effects on Cesium Diffusion in H-451 Graphite", Proc . 16th Biennial Conference on Carbon: Extended Abstracts and Pro- gram, held at San Diego CA, July 18-22, 1983, University of California San Diego, pp. 556-557. 109. D. MÜLLER and R. MOORMANN, "Auswirkungen einer Adaption des Fallenmodells auf die Cs-Quellterme bei Kernaufheizung im HTR- Modul", Technical Note KFA-ISF-TN-16/90 11, Research Center Jülich (1990). 110. K. VERFONDERN et. al., "Sensitivitdtsstudien zur Spaltprodukt/Graphit-Wechselwirkung für Unfallabläufe mit Kernaufheizung mittlerer HTR", KFA/ISF Status Seminar on Spaltprodukt/Graphit-Wechselwirkungen im HTR, held at Jülich, FRG, March 15, 1989 . 111. J. HERION, "Der EinfluB der Wandmaterialien auf die Ablagerung von Cs in HTR-Kühlkreisläufen", Jül-1155, Research Center Jülich (1975). 112. T. B. LINDEMER and T. M. BESMAN, "Chemical Thermodynamic Representation of UO2+x", J. Nucl. Mat. 130 (1985), pp . 473-488. 113. H. J. MATZKE, "Gas Release Mechanisms in UOz - A Critical Re`Tiew", Radiation Effects 53 (1980), pp. 119-242.

7.0 REFERENCES 95 11.4. G. FALTA and G. REITSAMER, "Freisetzung fester Spaltprodukte aus bestrahlten HTR- Brennstoffkernen bei Raumtemperatur nach Abtrennung der Teilchenbeschichtung", Part 1 : CH-262/80, Austrian Research Center Seibersdorf (1980). Part 2: CH-263/80, Austrian Research Center Seibersdorf (1980) . 115. G. MEISTER, "COROX, ein Rechenprogramm zur Simulation von Wasser- und Lufteinbrüchen in HTR", Private Communication, December 1990. 116. W. BREITUNG, "Oxygen Self and Chemical Diffusion Coefficients in U02±x", J. Nucl. Mat. 74 (1978), pp. 10-18 . 117. W. C. KUHLMANN, "Treatment of U02 with Water Vapor at High Temperatures", MCW-103, (1948). 118. J . T. BITTEL, L. H. SJODAHL, and J. F. WHITE, "Steam Oxidation Kinetics and Oxygen Diffusion in U02 at High Temper- atures", J . Am . Ceram. Soc. 52 (1969), pp. 446-451 . 119. J. B. LEWIS, "Fundamental Aspects of Defective Nuclear Fuel Behaviour and Fission Product Release", J . Nucl. Mat. 160 (1988), pp . 201-217. 120. W. SCHENK, D. PITZER, and H. NABIELEK, "Fission Product Release Profiles from Spherical HTR Fuel Elements and Accident Temperatures", Jül-2234, Research Center Jülich (1988) . 121 . K. SHIBA, "Fission Iodine and Xenon Release from the U02-U 308 System with Em- phasis on Radiation Damage", J . Nucl. Mat. 57 (1975), pp. 271-279. 122. W. SCHENK, R. MOORMANN, and H. NABIELEK, "Influence of Water Ingress on the Fission Product Release from HTR Fu- ei" Proc. Jahrestagung Kerntechnik '85, held at Karlsruhe, May, 5-7, 1992, Inforum GmbH, Bonn (1992), pp. 183-186. 123. R. MOORMANN, "Untersuchungen zum chemischen Verhalten der Spaltprodukte bei HTR-Kernaufheizstörfällen", Proc. Jahrestagung Kerntechnik '85, held at München, May, 21-23, 1985, Inforum GmbH, Bonn (1985), pp . 163-166 . 124. GULF GENERAL ATOMIC, "Public Service Company of Colorado 330 MW HTGR R&D Program", Quarterly Progress Report for Period Ending Sept . 30, 1966, GA-7453, Gulf General Atomic (1967), pp. 127-128 .

125. R. MOORMANN and K. PETERSEN, "REACT/THERMIX, ein Rechenprogramm zur Ermittlung der st6rfallbedingten Graphitkorrosion in Hochtemperaturreaktoren", Jü1-1782, Research Center Jülich (1982) . 126. G. IVENS and M. WIMMERS, "Der AVR als Testbett für Brennelemente", VDI-Berichte 729 (1989), pp . 161-176 . 127. N. INIOTAKIS and C. B. VON DER DECKEN, "Radiological Consequences of a Depressurization Accident Combined with Water Ingress in an HTR Modul-200", Nucl. Eng. Des. 109 (1988), pp. 299-306. 128. R. BORKOWSKI, H. BUNZ, and W. SCHÖCK, "Resuspension of Fission Products During Severe Accidents in LWR's", KfK-3987, Nuclear Research Center Karlsruhe (1985). 129. ORNL, "GCR Program", Semiannual Progress Report for Period Ending Sept. 30, 1966, ORNL-4035, Oak Ridge National Laboratory (1967), pp. 124-126. 130. ORNL, "GCR Program", Annual Progress Report 1973, ORNL-4975, Oak Ridge National Labora- tory (1975), Chap. 13. 131. F. P.O. ASHWORTH, "Information from Water Ingress Accident on AVR", JUI-Spez-43, Research Center Jülich (1979) . 132. W. C. HINDS, "Aerosol Technology", Wiley, New-York (1982), Chap. 6. 133. U. WAWRZIK, P. BIEDERMANN, and H. F. OETJEN, "Staub im AVR-Reaktor; Verhalten bei transienten Str6mungsbedin- gungen", Proc. Jahrestagung Kerntechnik '88, held at Travemünde, May, 17-19, 1988, Inforum GmbH, Bonn (1988), pp. 171-174. 134. G. SKRYME, "Attachment of Gaseous Fission Products to Aerosols", Proc. IAEA Specialists Meeting on Fission Product Release and Transport in Gas-Cooled Reactors, held at Berkeley,'UK, Oct. 22-25, 1985, IAEA IWGGCRj13, Vienna (1986), pp. 262-274. 135. J . A. GARLAND et. al., "Experimental Study of the Deposition of Iodine and Other Fission Pro- ducts in a Coolant Circuit of a CAGR", Proc. 5th Int. Meeting on Thermal Nuclear Reactor Safety, held at Karlsruhe/FRG, Sept. 9-13, 1984, KfK-3880, Nuclear Research Center Karlsruhe (1984), pp. 1423-1431 .

7.0 REFERENCES 97 136. GULF GENERAL ATOMIC, "Public Service Company of Colorado 330 MW HTGR R&D Program", Quarterly Progress Report for Period Ending Sept. 30, 1967, GA-8270, Gulf General Atomic (1967), pp. 141-143. 137. GULF GENERAL ATOMIC, "Public Service Company of Colorado 330 MW HTGR R&D Program", Quarterly Progress Report for Period Ending June. 30, 1968, GA-8725, Gulf General Atomic (1968), pp. 99-101 . 138. R. J. BLANCHARD, A. BROS, and J. GENTIL, "Out-of-pile Helium Loop for Liftoff Experiments", Proc. IAEA Specialists Meeting on Coolant Chemistry, Plate-out and De- contamination in Gas-Cooled Reactors, field at Jülich/FRG, Dec. 2-4, 1980, IAEA IWGGCR/2, Vienna (1981), pp. 19-25 . 139. F. F. DYER et. al., "Distribution of Radionuclides in the Peach Bottom HTGR Primary Cir- cuit During Core 2 Operation", ORNL-5188, Oak Ridge National Laboratory (1977) . 140. N. INIOTAKIS et. al., "Verhalten von Spaltprodukten bei der Simulation von Druckent- lastungsstörfällen im Kreislauf SMOC", Proc. Jahrestagung Kerntechnik '83, held at Berlin, June 14-16, 1983, Deutsches Atomforum, Bonn (1983), pp. 219-222. 141. R. GRIEMERT, "Staubablagerung in Gebieten abgelöster Strömung", Ril-1985, Research Center Jülich (1985) . 142. W. A. STARK, JR., "Cesium Solubility, Diffusion and Permeation in Zirconium Carbide", J. Nucl. Mat. 73 (1978), pp. 169-179. 143. T. OGAWA and K. IKAWA, "Diffusion of Metal Fission Products in ZrCl.0", J. Nucl. Mat. 105 (1982), pp. 331-334. 144. A. S. CHERNIKOV et. al., "Fission Product Diffusion in Fuel Element Materials for HTGR", Proc. IAEA Specialists Meeting on Fission Product Release and Transport in Gas-Cooled Reactors, held at Berkeley/UK, Oct. 22-25, 1985, IAEA IWGGCR/13, Vienna (1986), pp. 170-181 . 145. K. VERFONDERN, "Rechnerische Untersuchung der Rüekhaltefähigkeit des Graphits Mr Jod im Core des HTR-Modul", Technical Note KFA-ISF 14/85 1, Research Center Jiilich (1985). 146. R. MOORMANN and W. KATSCHER, "Examination on Graphite 'Burning' in HTGR's", 2nd Int. Symposium on Nuclear Reactor Severe Accident Chemistry, held at Toronto/Canada, June 5-9, 1988 .

147. NUKEM Internal Report NUKEM-83007 (1983). 148. H. U. BRINKMANN et.al, "Burner Off-Gas Examination in the Volume Reduction Experiment on Spent HTR-Fuel at AEE Winfrith", Proc. Jahrestagung Kerntechnik '86, held at Aachen, April 8-10, 1986, Deutsches Atomforum, Bonn (1986), pp. 313-316 . 149. G. H. LOHNERT, T. KINDT, and W. SCHERER, "Das Verhalten des HTR-Modul bei Reaktivitätsst6rfällen", Proc. Topp . Session on Stand der HTR-Sicherheitsforschung at the Jahrestagung Kerntechnik '89, held at Düsseldorf, May 9-11, 1989, Info- rum GmbH, Bonn (1989), pp. 26-47. 150. R. MOORMANN, "Source Term Estimation for Small-Sized HTR's", Jül-Report in preparation . 151. B. FORUTANPOUR and B. ROOS, "FIPERX, A FORTRAN V Program for the Solution of One-Dimensional and Non-Linear Diffusion Problems", GA Document No. 9904, Gulf General Atomic (1969) . 152. W. W. HUDRITSCH, "Fission Product Transport Code TRAMP", GA-D15190, General Atomic Company (1981). 153. R. MOORMANN and K. REISCHLE, "FALLDIF: Ein Rechenprogramm zur L6sung der durch Fallen gestörten Diffusionsgleichung", Technical Note KFA-ISF 14/90 II, Research Center Jülich (1990). 154. H. J. ALLELEIN, "Spaltproduktverhalten - Speziell Cs-137 - in HTR-TRISO-Brennstoff teilchen", J91- 1695, Research Center Jülich (1980). 155. B. F. MYERS, "Cesium Diffusion in Silicon Carbide During Post Irradiation Anneals", Technical Note KFA-HBK-TN-011`84, Research Center Jülich (1984). 156. K. VERFONDERN and D. MÜLLER, "Modeling of Fission Product Release Behavior from HTR Spherical Fuel Elements Under Accident Conditions", Proc. IAEA Specialists' Meeting on Behaviour of GCR Fuel under Acci- dent Conditions, held at Oak Ridge TN, Nov. 5-8, 1990, IAEA, IWGGCR/25, Vienna (1991), pp. 45-54.

7.0 REFERENCES 99 157. E. HOINKIS, "The Determination of Diffusion Coefficients of Cesium and Silver by the Release Method in As-Received, Oxidized and Neutron Irradiated Graphitic Matrix", Proc . Colloquium on Transport of Fission Products in Matrix and Graphite, held at Berlin/FRG, Nov. 9-11, 1981, HMI-B 372, Berlin (1983), pp. 77-102. 158. W. HENSEL and E. HOINKIS, "The Diffusion of Sr in the Graphitic Matrix A3-3 in Vacuum and in the Presence of Hydrogen", J . Nucl . Mat. 184 (1991), pp. 88-96. 159. K. VERFONDERN and D. MÜLLER, "FRESCO-1, Programm zur Berechnung der Spaltproduktfreisetzung aus dem Core eines Kugelhaufenreaktors bei Kernaufheizung Benutzerhandbuch", Internal Report KFA-ISR-IB-2J91, Research Center Jülich (1991) . 160. F. J. SANDALLS and M. R. WALFORD, "Laboratory Determinations of Strontium Diffusion Coefficients in Graphite", AERE-R 6911, Harweil (1972) . 161. B. F. MYERS, "Effect of Water Vapor on the Release of Fission Gases from Uranium Oxicarbide in High-Temperature, Gas-Cooled Reactor Coated Fuel Parti- cles", J. Am . Ceram. Soc. 75 (1992), pp . 686-693. 162. K. MINATO, "Diffusion Coefficients of Fission Products in U02, PYC, S'C, Graphite Matrix and IG-110 Graphite", IAEA Technical Workshop on Unification of Coated Particle Performance Models and Fission Product Transport Data for the HTR, held at Jülich, FRG, Dec. 2-4, 1991 . 163. D. T. GOODIN and H. NABIELEK, "The Performance of HTR Fuel in Accidents", Technical Note KFA-HBK-TN-19%85, Research Center Jülich (1985) . 164. B. F. MYERS and W. E. BELL, "Cesium Transport Data for HTGR Systems", GA-A13990, General Atomic Company (1979) . 165. B. F. MYERS and W. E. BELL, "Strontium Transport Data for HTGR Systems", GA-A13168, General Atomic Company (1974) .

166. W. AMIAN et. al., "Results of Fission Product and Actinide Studies in Coated Fuel Particles and Matrix Graphite", Proc. BNES Conference on Gas-Cooled Reactors Today, held at Bristol/UK, Sept. 20-24, 1982, British Nuclear Energy Society, London (1982), pp. 153-160 . 167. K. HAYASHI and K. FUKUDA, "Diffusion Coefficients of Fission Products in the UOz Kernel and Pyrocarbon Layer of BISO-Coated Fuel Particles at Extremely High Tem- peratures", J . Nucl. Mat. 174 (1990), pp . 35-44. 168. W. AMIAN and D. ST6VER, "Diffusion of Silver and Cesium in Silicon-Carbide Coatings of Fuel Parti- cles for High-Temperature Gas-Cooled Reactors", Nucl. Techn. 61 (1983), pp. 475-486. 169. H. NABIELEK, R. GONTARD, and B. F. MYERS, "I . Version of Joint KFA/GA Fission Product Data Book", Internal Report KFA-HBK-IB-11/81, Research Center Jülich (1981). 170. D. ALBERSTEIN, P. D. SMITH, and M. J . HAIRE, "Metallic Fission Product Release from the HTGR Core", GA-A13258, General Atomic Company (1975). 171. H. NABIELEK, "Fission Product Data from the British HTR Work", Colloquium on Transport of Fission Products in Matrix and Graphite, held at Berlin/FRG, Nov. 9-11, 1981, Paper C8 (not published in pros.). 172. P. ZOLLER, "Das Transportverhalten der Spaltprodukte Cäsium und Strontium in beschichteten Brennstoffteilchen für Hochtemperatur-Reaktoren unter Bestrahlungsbedingungen", Jill-1324, Research Center Jülich (1976) . 173. K. HAYASHI and K. FUKUDA, "Diffusion Coefficients of Cesium in Un-Irradiated Graphite and Compar- ison with Those Obtained from In-Pile Experiments", J . Nucl. Mat. 168 (1989), pp. 328-336. 174. R. A. CAUSEY, "Silver Transport in H-451 Graphite", Trans . Am . Nucl. Soc. 38 (1981), pp . 316-317 . 175. D. CUBICIOTTI, "The Diffusion of Xenon from Uranium Carbide Impregnated Graphite at High Temperatures," NAA-SR-194, North American Aviation (1952). 176. P. E. GETHARD and L. R. ZUMWALT, "Diffusion of Metallic Fission Products in Pyrolytic Carbon", Nucl. Appl. 3 (1967), pp. 679-685.

7.0 REFERENCES 10 1 1.77. B. F. MYERS, "Selected Aspects of Fission Product Behavior in High-Temperature, Gas- Cooled Reactors", IAEA Technical Workshop on Fission Product Transport Data for the HTR, held at Rilich, FRG, March 26-27,1992. 178. K. FUKUDA, K. HAYASHI, and K. SHIBA, "Fuel Behavior and Fission Product Release under HTGR Accident Con- ditions", in: J. T. ROGERS (ed.), "Fission Product Transport Processes in Reactor Accidents", Hemisphere Publishing Corporation/USA (1990), pp. 197-204. 179. HOBEG, "Arbeiten der HOBEG",Abschlußberichtease under HTGR FuE - 83007, Final Report for Period July 1st, 1977 to Dec. 31st, 1981, Hanau/FRG (1983) . 180. M. T. MORGAN and A. P. MALINAUSKAS, "Cesium Release and Transport in BISO-Coated Fuel Particles", Nucl . Techn. 35 (1977), pp. 457-464. 181. K. HILPERT, H. GERADS, and D. KOBERTZ, "Sorption of Strontium by Graphtic Materials", Ber. Bunsenges. Phys. Chem. 89 (l985), pp . 43-48. 182. J. KWASNY, K. HILPERT, and H. NICKEL, "Cesium-Sorptionsuntersuchungen an graphitischen Reaktorwerkstoffen", Jü1-2353, Research Center Jülich (1990) . 183. GENERAL ATOMIC COMPANY, "HTGR Fuels and Core Development Program", Quarterly Progress Report for Period Ending August 31, 1977, GA-A14479, General Atomic Company (1977). 184. R. MOORMANN and R. KOSCHMIEDER, "Versuch einer verbesserten formelmäßigen Erfassung der Cs- und Sr- / A3-3 Sorption im Freundlich-Bereich", Technical Note KFA-ISF 4/88 11, Research Center Jülich (1988) . 185. K. VERFONDERN, K. HILPERT, and R. MOORMANN, "Sorption of Fission Products on Graphite and its Influence on their Re- lease Behavior in a Pebble Bed HTR under Accident Conditions", Proc. IAEA Specialists Meeting on Fission Product Release and Transport in Gas-Cooled Reactors, held at Berkeley/UK, Oct. 22-25, 1985, IAEA IWGGCR/13, Vienna (1986), pp . 371-384. 186. A. A. GUSEV et. al., "Experimental Data and Calculations to Analyze the Factors of HTGR Fuel Elements Service Life", Proc. 1st Soviet/German Seminar on Fuel and Graphite for HTR, held at Moscow, Russia, July 23-27, 1990, KFA Internal Report KFA-HTA-IB-6/90, Research Center Rilich (1990), pp. 455-472.

187. 1 . G. DEGALTSEV et. al., "Postirradiation Examination of Particles and Fuel Elements", Proc. 1st Soviet/German Seminar on Fuel and Graphite for HTR, held at Moscow, Russia, July 23-27, 1990, KFA Internal Report KFA-HTA-113-6/90, Research Center Jülich (1990), pp. 193-221 . 188. K. HILPERT et. al., "Sorption of Cesium and its Vaporization from Graphtic Materials at High Temperatures", High Temp. High Press. 20 (1988), pp . 157-164 . 189. A. A. KHRULEV et. al ., "State of Works on Fuel Elements and Fission Products Transport for HTGR Radiation Safety Provision", IAEA Technical Workshop on Unification of Coated Particle Performance Models and Fission Product Transport Data for the HTR, held at Jülich, FRG, Dec. 2-4,1991 . 190. R. A. LORENZ, F. F. DYER, and R. L. TOWNS, "Sorption/Desorption Behavior of Iodine on Graphite", ORNL-TM-8284, Oak Ridge National Laboratory (1982). 191. E. PROKSCH, A. STRIGL, and H. NABIELEK, "Production of Carbon Monoxide During Burnup of U02 Kerneled HTR Fuel Particles", J. Nucl. Mat. 107 (1982), pp. 280-285 . 192. A. STRIGL, G. PESCHTA, and E. PROKSCH, "CO-Messungen an Mischoxidteilchen", CH-328/84, Austrian Research Center Seibersdorf (1984) . 193. F. J. HOMAN et . al., "Stoichiometric Effects on Performance of High-Temperature Gas-Cooled Reactor Fuels from the U-C-0 System", Nucl. Techn. 35 (1977), pp. 428-441 . 194. D. T. GOODIN, H. NABIELEK, and W. SCHENK, "Accident Condition Testing of US and FRG High-Temperature Gas- Cooled Reactor Fuels", JOI-Spez-286, Research Center Jülich (1985) and GA-A17820, GA Technologies (1985) . 195. F. C. MONTGOMERY, K. E. PARTAIN, and H. STALEY, "Final FY 82 Report on Fission Product - SiC Reactions," GA Document No. 906641, General Atomics (1982) . 196. H. NABIELEK et. al., "Development of Advanced HTR Fuel Elements", Nucl. Eng. Des. 121 (1990), pp. 199-210.

7.0 REFERENCES 103 104

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL

The classical approach for a diffusion model is the numerical solution of the Fickian equation using effective diffusion coefficients for the fission product species in different kinds of reactor materials. "Effective" means that all possible transport mechanisms are summarized in a simplified single transport process. The diffusion model is generally regarded to be valid both for normal operation and for accident conditions.

The diffusion coefficients are usually given as an Arrhenius type equation as a function of temperature. Empirically found deviations from this behavior could be overcome by assuming additional dependencies, for instance on fast neutron fluence 34 or fission product concentration or burnup, or by combining diffusion processes with different activation energies in different temperature ranges:

D(T,I",c, . . .) -.

where Do : pre-exponential factor [m'/s] Q : activation energy [J/moll T : temperature [K] r: fast neutron fluence [1025 m-2 , E>0 .1 McV3 -3 3 c : fission product concentration [m R : gas constant, R = 8 .3143 [J/(mol K)]

Diffusion coefficients have been adopted from the evaluation of numerous irradiation and heating experiments with complete spherical fuel elements, fuel compacts, single fuel particles, or graphite samples.

e.g., a different cutoff limit 14 Different units for the fast neutron fluence are found in the literature; . correct unit is explicitly for the neutron energy E > 0.1 MeV, E > 0.18 MeV, or ED\' The mentioned whereever necessary.

105 Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL Recommendations of transport data have been collected over many years and were continuously revised. A set of data to be used for predicting fission product transport during HTGR normal operation has been established in a joint effort of KFA and German industries within the project "Hochtemperaturreaktor- Brennstoffkreislauf" (HBK). Its purpose was to summarize the data base for HTGR design calculations for licensing procedure. The HBK data set was last published in 1986 (Ref. 54).

For fission product transport studies under accident conditions, however, additional or new recommendations have been derived from corresponding experimental data to meet the requirements of HTGR safety analyses. An extended set of transport data was given in the status report (Ref. 1) .

An observed high permeability for a fission product species in a material zone is simulated in a diffusion model by choosing a comparably "large" diffusion coefficient. A typical example is the buffer layer. In contrast, a complete retention can be simulated by a very "small" diffusion coefficient; e .g., for fission gases in intact particle coating layers.

A common assumption in most diffusion models is the neglect of any effects of sorption or trapping within a coating layer or of any preferential retention in a specific layer. The ratio of fission product concentrations at the surfaces of contiguous materials ("partition coefficient") is assumed to be one; consequently, the concentrations are equal at contiguous surfaces. Measurements of this ratio range from about 0.3 to 3 and may change with temperature and fast fluence (Ref. 43). For lack of adequate measurements of the ratio under a sufficiently large range of conditions, the error associated with the assumed value of 1 .0 is accepted . The few experimental studies related to desorption from coating materials indicate some kind of discontinuity in concentration profiles at coating interfaces, for instance for cesium between buffer and iPyC layer or the significant palladium peaks at the inner surface of the SiC layer (Ref. 44) in micro- probe profiles. The increase of cesium concentration towards the surface of Soviet coated particles was explained by a contamination of the outer coating layer during the disintegration of the fuel sphere (Ref. 46).

Experiments at KFA have been proposed in 1989 for investigation of the thermochemical partition coefficients at material boundaries to better estimate the magnitude of this effect . The development of an alternative calculation model

106

is underway at KFA which considers the fission product transport to be dependent on the gradient of the chemical potential rather than of the concentration (Ref. 45). The modeling differences, however, are not expected to be important for small-sized HTGRs under accident conditions since the enormous retention capability of colder graphite regions in the core - described by a partition coefficient (sorption isotherms) at the graphite/coolant boundary - keeps the release of metallic fission products negligibly small.

Tables A-1 to A-VI and Figs. A-1 to A-6, respectively, try to summarize the actual knowledge in transport data and attempt to compare the results from different countries. Differences from the former status report (Ref. 1) are explicitly mentioned.

A.1 FUEL KERNEL

Diffusion coefficients in U02 fissile particle kernels which were used in recent German safety analyses have been derived from irradiation and heating experiments with designed-to-fail U02 particles (kernel + buffer layer) . According to an evaluation of the German irradiation and heating experiment FRJ2-P28, cesium and iodine release under accident conditions is considered at KFA to be best approximated by a two-branch diffusion coefficient while those for strontium and silver were modified by a factor of 20 and 10, respectively, compared to the recommendation of the last HBK data set. A different recommendation is given for iodine at lower temperatures (c 1000 °C) which was derived from irradiation experiments in the R2 reactor at Studsvik (Ref. 77) . There is no change of KFA data compared to the status report (Ref. I) .

In US, most data were derived from Th02 fertile kernels . A significant dependence of the reduced diffusion coefficient on temperature and burnup but no dependence on fast fluence was found. The diffusion data for metallic fission products (Ref. 161) are also used for U02, with a separate set of constants used for UCO kernels which are slightly higher than those for U02. UC2 fuel is assumed to retain no metallic fission products. However, for safety analysis calculations, the use of KFA data for the metallic fission products is recommended .

Appendix A. TRANSPORT DATA FOR DIFFUSION :MODEL 107 Other US data recommended for cesium (Ref. 164) and strontium (Ref. 165) represent a fit through many data from different authors.

Japanese transport data are identical for cesium and silver . Both curves are above corresponding German and US data.

The diffusion data for UO2 are summarized in table A-I and plotted in Fig. A-l . Data and plots for (Th,U)O2 fuel kernels mainly used in BISO particles are not presented here; for further detail see Ref. 1 .

The assumption of a defective or failed particle in the diffusion code FRESCO-I1 is equivalent to simulating an exposed particle kernel. This means that fission products released from the particle kernel by diffusive transport are immediately released to the surrounding graphite structure. The failure of a particle coating, however, is not necessarily equivalent to a non-existing coating; such an assumption is a conservative approach. There is experimental evidence for a better retention of fission products in particles with a broken coating compared to exposed kernels. The simulation of a partly existing coating as a kind of a 3rd type of particle (besides intact particles and exposed kernels) has been proposed by JAERI (Ref. 35, see also later section on silicon carbide). Another possibility of modeling a broken but still existent coating is the choice of a smaller diffusion coefficient in the kernel as an "effective" diffusion coefficient for this 3rd type of particle as has been done in the FRESCO-I1 code (Ref. 75) . A postcalculation of the 1800 °C isothermal heating test HFR-K3/3 has shown that the reduction of the kernel diffusion coefficient by two orders of magnitude simulates a failed but still existing coating and reproduces the high-release peak of the observed IMGA bimodal distribution of single particle inventories (Ref. 70) .

A.2 PYROGARBON

Diffusion coefficients in pyrocarbon were mostly derived from early experimental data with LTI- and HTI-PyC from BISO particles. Due to its comparably smaller retention capability, diffusion in PyC has not been investigated in that detail as it has been for the main barrier silicon carbide. At temperatures beyond

108

1900 °C, it is recommended to assume rapid diffusive transport (similar to the buffer layer) for metallic fission products in LTI-PyC. For HTI-PyC, this critical temperature has been assumed in German safety analysis calculations for the THTR-300 to be as low as 1200 °C, except cesium for which experimental evidence suggests much better retention than strontium or silver. The diffusion coefficient of krypton in PyC is based on an evaluation of GA heating data at 2050 °C and ramp tests and of KFA 1600 °C tests (Ref. 163). No modification of KFA transport data is considered to take into account possible irradiation damage in PyC due to a high probability of irradiation damage annealing (Ref. 163) . No change of KFA data has been made compared to the status report (Ref. 1) .

US recommendations for strontium and silver as well as for iodine and fission gases are identical to those at KFA. The data for cesium (Ref. 164) and strontium (Ref. 165) represent an average of the measurements of many authors. Early US studies (Ref. 176) have also found the cesium transport data through isotropic pyrocarbon to be orders of magnitude lower than those for strontium, both being in the range of the later reported data.

In Japan, several studies have been made on cesium in different temperature ranges (Ref. 162). The Japanese xenon data are about two orders of magnitude below the KFA krypton data.

Diffusion data given by other authors (Refs. 46, 144, 166, 176) are fairly close to each other despite their strong dependence on the pyrocarbon deposition conditions during manufacture and material characteristics. This dependence was reported to be the reason for the relatively large difference in JAERI and KFA activation energies for cesium (Ref. 167).

plotted in The diffusion data for pyrocarbon are summarized in table A-II and particles are not Fig. A-2. Data and plots for HTI material mainly used in BISO presented here; for further detail see Ref. 1 .

model to have no sig- The carbonaceous buffer layer is assumed in the diffusion calculations, a value nificant retention capability for fission products. In FRG M2 dependence (acti- of the diffusion coefficient of 10,$ /s with no temperature recommendation for the corre- vation energy equal to zero) is used. The US 2 is negligible since either sponding diffusion coefficient is 10 10 In /s; the difference

MODEL 109 Appendix A. TRANSPORT DATA FOR DIFFUSION one is significantly larger than the diffusion coefficient for the other coating materials.

A.3 SILICON CARBIDE

The transport properties of fission products in silicon carbide, as the most efficient barrier of the coated particle, were the subject of many detailed studies in the past years. The experimental data gave impetus to either confirm and/or re- fine existing models or even to develop new ones.

The original way to describe the permeability of the SiC coating for fission products was a diffusion coefficient derived from numerous heating tests at KFA with single coated fuel particles . The radionuclide cesium was investigated more than any other relevant metallic fission product species. The Arrhenius relation for cesium proposed by Allelein (Ref. 154) entered the HBK data set as recommended data for use in HTGR core release predictions under normal operating conditions and was recommended in Ref. 1 .

For irradiation experiments with modern HTGR fuel, the Allelein diffusion coefficient was found to overestimate in some cases the cesium release fraction from fuel particles. An evaluation of the experimental data by Christ (Ref. 48) has taken the fast neutron fluence as a parameter for the Arrhenius relation, thus influencing the amount of cesium penetrating the SiC layer during irradiation and the delay of the diffusive break-through at elevated temperatures. This new recommendation for the low-temperature branch of KFA data is - for zero fast fluence - lower- compared to the Allelein data but will come close to it with increasing fast fluences (Fig . A-3b) .

A theoretical evaluation of several researchers' experimental cesium release data from coated particles was made by Myers in the early 1980s. The result was a diffusion coefficient subdivided into a low temperature branch and a high temperature branch dominant at > 1600 °C. The data at high temperatures were likewise subdivided and identified to characterize two different types of silicon carbide material (Ref. 155) . The upper curve (Fig. A-3b) was chosen in 1986 as

the reference data for accident temperature conditions with the goal of being at least conservative in German safety analyses calculations (Ref. 1) .

Experimental results from heating tests in the KiiFA furnace at KFA with modern HTGR spherical fuel elements beginning in 1984 (Ref. 120) however, showed a discrepancy in the high temperature release behavior from single coated particles and from complete fuel balls, with the latter exhibiting much lower release fractions. Due to a considerable retentivity of the matrix graphite for metallic fission products, due to the optimized fuel manufacturing process which may have influenced the fission product transport characteristics in the particle coating, and due to the better statistics of about 104 particles per fuel element, experiments with single particles were no longer regarded as representative of modern high quality fuel .

This experience coincides with the fact that postcalculations of cesium release from heated fuel elements using the (upper) Myers' diffusion coefficient have led to an overestimation by up to several orders of magnitude. A recent evaluation of data from all heating experiments conducted in the KdFA furnace so far - a total of 44 tests with respect to cesium - using the diffusion code FRESCO-11 has led to a new KFA recommendation for a diffusion coefficient in silicon carbide (Ref. 156) which is very similar to the lower Myers curve.

An analogous evaluation has been made for strontium on the basis of 11 KdFA heating tests with modern HTGR fuel in the temperature range between 1600 and 1800 °C. The new diffusion coefficient (Ref. 156) shown in Fig. A-3a has a higher activation energy resulting in a diffusion coefficient lower by a factor of about 20 at 1600 °C compared to the old recommendation. Unfortunately, the combination of this new high-temperature Arrhenius relation with the old HBK relation for normal operation temperatures into a single two-branch diffusion coefficient significantly underestimates the experimental retention quality of SiC for strontium at 1600 °C (Ref. 156) . In order to benefit from this potential for reduced release, more strontium release data at lower temperatures should be made available.

KFA silver transport data have been taken from the HBK data set. No evaluation of available silver release data from KOFA heating tests has been made so far because no consistent release behavior was observed and is in many cases not reproducible by any existing model . A possible explanation as proposed by

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL III Myers (see section 2.3.2) could be the silver to be trapped at neutron-induced defects in the silicon carbide structure at normal operation temperatures resulting in a burst release of the trapped silver at elevated temperatures.

According to the definition of the "integrated Failure and Release Model for Standard Particles" which represents the US reference model (Ref. 8), no diffusion of cesium (and strontium) through silicon carbide is assumed to occur. There are, however, US diffusion data available which were published before the statistical model was created (Refs. 155, 169, 170) . Ongoing work by Martin, based on Ref. 36, suggests that a fast-fluence-dependent diffusion coefficient similar to that of Christ (Ref. 48) represents an improvement over the reference model.

A surprising result as mentioned in Ref. 162 is given with the Japanese data on xenon diffusion in silicon carbide for the high-temperature branch (> 1400 °C) being close to the cesium data at temperatures > 1800 °C (Fig. A-3a). No big difference between Japanese and KFA data was found for cesium and silver. For strontium, the same data as KFA for the low-temperature branch are used (Fig. A-3a).

The uncertainty range for transport data in silicon carbide is a result of the very low release level for high quality fuel at temperatures S 1600 °C which sometimes only allows the indication of an upper limit. When considering heating tests with complete fuel elements, other parameters such as heavy metal contamination fraction in the graphite gain importance (see also the following section A.5).

The approach in the JAERI code FORNAX which uses a 3rd type of particle with a degraded SiC layer simulated by larger values for the diffusion coefficient has been previously mentioned ; no specific transport data have been published so far.

R611ig has recently provided an interesting analysis of the Cs-137 release data for FRG sphere R2-K13/'l heated to 1600 °C for 1000 h (Ref. 74) . Nearly 2000 particles were gamma-counted, and 7 % had released over 10 % of their Cs-137 inventory, with a small number of particles releasing 50 to 90 %. Röllig applied a diffusion model to the release data, but rather than a single diffusion coefficient he assumed a log-normal distribution of values around the average diffusion coefficient. He obtained reasonable agreement with the data for particle release as a function of position within the sphere. This method of assuming a distribution

of diffusion coefficients is a promising approach to account for the inherent microstructural variation in SiC and its effect on particle-to-particle release com- parisons .

A.4 GRAPHITE

The uncertainty for the given transport data in graphitic materials is considered to be fairly high. The metallic radionuclide transport in graphite is strongly dependent on graphite type and nature, structure, temperature, fission product concentration, state of oxidation, irradiation damage, coolant gas presssure, and interference with other fission product species. The trapping mechanism is not considered, which is known to better approximate the real transport process in graphite. The overall uncertainty range is estimated to be at least one order of magnitude, for silver as high as three orders of magnitude. For safety analysis purposes, however, transport data for the oversimplifying Fickian diffusion model in graphite can be used to obtain realistic, or at least conservative radionuclide release data .

A.4.1 Matrix Graphite

Effective diffusion data are relatively well known for cesium and silver (Ref. 157) and strontium (Ref. 158) in German A3-3 matrix graphite in the Henrian concentration regime as well as the influence of irradiation and corrosion on the diffusive process. The matrix graphite type A3-3 was used for the THTR-300 spherical fuel elements and was chosen to be the reference fuel element graphite for future German HTGRs. In contrast, all of the modern German fuel elements which were used in irradiation and postirradiation experiments consisted of the matrix graphite type A3-27. Comparative measurements of Hoinkis showed diffusion coefficients for cesium and silver in A3-27 lower by a factor of 20 and 7, fission pro- respectively, at 1000 °C (Ref. 157). The KFA data for the metallic

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL 113 ducts given in table A-IVa and shown in Fig. A-4a refer to irradiated A3-3 matrix graphite, the same as in the status report (Ref. 1). 15

According to the experience from measurements, cesium and strontium diffusion through graphite is assumed in German safety analyses to proceed rapidly at temperatures beyond 2000 ° C, for silver already at temperatures > 1000 °C.

The evaluation of cesium inventory measurements in the matrix of heated fuel elements with the KFA code FRESCO-11 have also demonstrated the diffusion coefficient in A3-27 to be at least one order of magnitude lower compared to A3-3 in the accident temperature range of 1600 to 1800 °C. But the fitted diffusion coefficients for A3-27 cannot consistently be described in a new Arrhenius relation on a lower level (Ref. 156). The choice of heavy metal contamination fraction in the matrix which is used as input data for the model calculations has a major effect on the release level from the fuel element, especially at lower heating temperatures S 1600 ° C (when the release from the coated particles is comparably small) and especially for strontium which is more strongly bound in graphite than is cesium.l 6 An adequate recommendation for cesium and strontium transport in A3-27 matrix graphite may be the reduction of the Hoinkis diffusion coefficient curves (for A3-3) by one order of magnitude which would still envelope the calculated results (Ref. 156) .

15 The difference in KFA cesium data between table A-lVa and Ref. 1 is due to an error in the original Hoinkis paper in Ref. 157. However, the irradiation dependence discussed in Hoinkis' paper was found to not significantly change the transport data. The difference in KFA strontium data between table A-IVa and Ref. I is due to a slight discrepancy between earlier reported data and the final publication (Ref. 158).

16 The input data for heavymetal contamination fraction and for the diffusion coefficients in SiC and matrix graphite are usually varied until the calculated release data fit the measurements. The adjustment of fuel element release to a lower level can - if the release from the coated particles is correctly reproduced - be achieved by reducing either the diffusion coefficient in graphite or the contamination fraction.. The latter is not explicitly known for every single fuel element, only average values can be estimated.

US diffusion data on fuel compact matrix materials are not given due to the assumption of a rapid transport of the metallic and gaseous fission products through this porous material zone (Ref. 8) .

Japanese data have been published for strontium in graphite matrix for the VHTR design (Ref. 63) which is one to two orders of magnitude above German reference data. The Japanese data are, however, somewhat smaller compared to early KFA measurements conducted for strontium and cesium in irradiated A3 matrix of AVR fuel elements (Ref. 172).

Other studies have been made with the British compacted natural grap'te matrix material (Ref. 171) which are fairly close to the corresponding German reference data.

Fission gases released from the coated particles are assumed in US modeling to be transported through the fuel compact matrix material without any time delay both under normal operating and accident conditions. This assumption has the same result as the German approach which uses a large diffusion coefficient for xenon in helium to simulate fission gas and iodine transport in the graphite pores (= grain boundaries). The US modeling of radionuclide transport in graphite has been recently refined by introducing the effect of graphite oxidation into the diffusion coefficient by making the pre-exponential factor a function of weight percent burnoff.

The transport behavior of iodine in graphite under accident conditions is considered in German model calculations to be similar to that of the noble gases krypton and xenon . Rather than simulating an effective, one-phase diffusion, the iodine transport is conservatively treated in the KFA reference modeling to consist of a slow diffusion phase out of the graphite grain proposed by Müller (Ref. 77) at temperatures S 1250 °C and of a rapid diffusion phase via the pores and the graphite grain boundaries. The iodine from the heavy metal contamination is assumed to be buried deep inside the grains. In contrast, iodine released from defective particles into the graphite is assumed to be immediately transported in

a combination of fired 17 The fuel compact matrix material is not a matrix graphite but of condensible carbonaceous material, graphite grains, and coated fuel particles. The transport release of fission products will be mainly on the surfaces of the carbonaceous material. The condensible fission products from the compact is by desorption.

Appendix A. TRANSPORT DATA FOR DIFFUSION :MODEL 11 5 the graphite pores. The quick phase transport is independent of temperature and corresponds to the diffusive behavior for xenon in helium, and does not allow for any realistic retention in the graphite .

The transport of iodine and fission gases in graphite has been recently the subject of a special series of KFA experiments (Ref. 82). Spherical fuel elements with low burnup were heated to investigate the release behavior of short-lived isotopes originating from the heavy metal contamination . Iodine release curves were found to be akin to those for xenon, with both supporting the interpretation of a release from traps in the matrix graphite. "Effective" diffusion coefficients derived from these experimental data using the FRESCO model (Ref. 156) revealed (see Fig. A-4b) that the reference diffusion coefficient as explained in the above paragraph was too small at lower temperatures and too large at higher temperatures > 1250 "C. But from the arrangement of the fitted data, an activation energy could be recognized which is similar to that for an effective diffusion coefficient for xenon transport in irradiated AUF graphite which was impregnated with uranium carbide (Ref. 175). This Canadian investigation had shown that relatively high fractions of xenon were retained in the graphite even at temperatures up to 2000 "C.

The above mentioned evaluation of the KFA heating experiments with the FRESCO-11 code was repeated, but now using for the contamination fraction in (1* the spheres a single value 10-7) rather than the release value measured at the end of each series . The new calculation results shown in Fig. A-6b are up to two orders of magnitude lower than the earlier results, and the low temperature data fall close to the Müller slow phase diffusion coefficient. However, all things considered, these calculational results do not seem to represent a good basis for recommending a new diffusivity. The old one remains sufficiently conservative for safety analyses, although an approach using a trapping mechanism will be the more accurate one.

A.4.2 Structural Graphite

No diffusion data are available for the German reference reflector graphite ASR-IRS. The US data for H-451 graphite grade have been used instead in model calculations . The transport of iodine and fission gases in structural

graphite is only considered via the rapid diffusive phase since they have penetrated the graphite from the coolant side, thus being transported using the pores rather than entering the grains.

US data for cesium and strontium which approximate various authors' measurements and are recommended to be used in HTGR core (graphite type: H-451) release predictions were presented in the table A-1Vb to also represent KFA transport data for structural graphite. The silver transport in H-451 measured at temperatures up to 800 °C was supposed in Ref. 174 to consist of a slow phase with high concentrations near the surface and of a fast phase with low concentrations away from the surface.

JAERI studies of IG-110 structural graphite in in-pile experiments have revealed significant differences in cesium and silver diffusion data relative to German and US data (Ref. 64). In particular, silver released from the fuel compact was found to be effectively retained in the graphite sleeve for low burnups < 2 °1o FIMA . Diffusion coefficients for cesium obtained from experiments with Cs-impregnated IG-110 graphite specimens were found to be larger by 3-4 orders of magnitude and with a lower activation energy compared to those data obtained from the in-pile experiments. An explanation is proposed by assuming an additional trapping effect by irradiation-induced surface defects with stronger binding energies than those of existing traps (Ref. 173) .

British diffusion data shown in Fig. A-5 refer to fuel tube graphite AGL-9 (Ref. 171) .

The diffusion data for structural graphite are summarized in table A-1Vb and plotted in Fig. A-4c.

A.4.3 Concentration Dependence of the Diffusion Coefficient

Experimental data on the concentration dependence of transport data have been collected by several authors (discussed in Refs. 165, 171). A significant effect was found for strontium where a tremendous increase of diffusivity by several orders of magnitude occurred as soon as a certain concentration was exceeded (Ref.

Appendix A, TRANSPORT DATA FOR DIFFUSION MODEL 117

160). Only small effects were found for silver and cesium. The concentration dependence of strontium diffusion is expressed in the empirical equation

log D (T, egr) = 4.51 - 2.58 " log cgr 29300 - 7000 -log cgr ) (A -- 2) T

where Cgr in the range 60 - 2000 [log Sr/g C]

The following empirical strontium diffusion coefficient dependent on temperature and concentration was found to be in good agreement with experimental data and recommended for US safety studies (Ref. 165) :

64000 " X 1 1 cg,*D(T, egr) = 3.47 " 10'"1s er R " ( T - 12 33 l (A -- 3)

where

4 0.5805 x = 0.422 3 C 1 + 80

and Cgr in the range 0 .7 - 50 [ymol. Sr/g C]

This diffusion coefficient is plotted in Fig. A-4d .

A modeling refinement for diffusive transport of cesium and strontium in graphite which simulates this effect uses the British experimental data obtained for strontium (Ref. 160). In the FRESCO model, corresponding sorption data for strontium and cesium (see appendix C) are taken to introduce a factor con- sisting of the ratio of the partition coefficient based on Freundlich sorption isotherms over that based on Henry sorption isotherms (Ref, 75). The partition coefficient a = a(T) is defined by the equation

c cgr (A -- 4)

where cgas : concentration in coolant [Atoms/m 31 cgr : concentration in graphite surface layer [Atoms/m3]

The concentration dependence is given by modifying the pre-exponential term of the diffusion coefficient in graphite by a factor which is equal to 1 for the Henrian concentration regime and greater than 1 for the Freundlich concentration regime :

aFreundlfch D cg,, _ (A _- 5) a ) Do 4Henry

where cgas = a * cgr

This type of concentration dependence is presented in Fig. A-4d for strontium reproducing the results of Ref. 160, and in Fig. A-4e for cesium. This procedure is also applied to the structural graphite of the reflectors in the 'core' version of FRESCO (Ref. 159) .

Another approach has been proposed in Ref. 41 which is employed in the SPTRAN code. The pre-exponential factor of the diffusion coefficient is additionally dependent on temperature

r " EF

(A - 6) Do ( T, Do

where ct : transition concentration between Henry and Freundlich regime (see Appendix C) empirical parameter parameter of the Freundlich sorption isotherm corresponds to E of equation (C-2) in Appendix C EF

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL 119 A value of 0.25 for the parameter r was fitted to best approximate cesium experimental data (Ref. 33). The diffusion coefficient according to equation (A-6) is shown in Fig. A-4e (dashed curves).

A.5 ZIRCONIUM CARBIDE

Zirconium carbide is considered to be a promising substitute for the SiC coating layer in the fuel particle. Several authors have investigated its potential for re- taining metallic fission products. In Ref. 142, the cesium diffusivity in ZrC was -18 estimated to be on the order of 10 to 10-16 M21s with an activation energy of 50 kJ/mol for the temperatures between about 1200 and 1600 °C. Japanese experiments (Ref. 143) indicate strontium and barium diffusion data at 1400 °C are in a similar range. Russian data (Ref. 144) on silver and barium on a higher temperature range are available.

The diffusion data for zirconium carbide are summarized in table A-V and plotted in Fig. A-5.

A comparison has been made in temperature ramp tests for the cesium release from Japanese TRISO-SiC and TRISO-ZrC fuel particles with approximately 4 FIMA burnup. The ZrC coated particle has demonstrated better cesium retention by up to one order of magnitude at temperatures S 2000 °C and up to three orders of magnitude at 2200 °C (Ref. 3) . Despite these interesting retention properties, the complete development of a ZrC coating interlayer as a substitute for SiC would require a comprehensive time- and money-consuming qualification program. Japan's HTTR program includes the development of a ZrC coating in advanced fuel for later cores (Ref. 5) . The evaluation of the US and Japanese experience with zirconium carbide coated fuel for a VHTR with gas turbine application is part of a recently published EPRI study prepared by the University of Tennessee in Knoxville (Ref. 47).

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL 121 w: r

ö N O N 00 V' uux M M N a

u N N N O N Q

ca r-~ C4 C a\ 00 kn 'c ,. aS 00 in r r- p 00 ~N~ c~ u ONIT o ä

N .C~_ 0000 M ON - - 00 M ON 10%.0 121, ` Zl Ü`i Qi Q., I-CN 1l00r rr N tnN %-0--00 C C

DU U CJ r ZQ Ü H a

C d -tt ÜciQ~ Üri1Q E- co6Q

122

10-4

FRG --- USA Japan

U0 2 ( based on kernel radius r = 254 )Am )

c 16-

0 v c 12-i0-14 - 0 0

a 10-16 -

10-18 -

~----- Temperature ['01 2100 1800 1600 1400 1000 800 10-20 -'-i ...~ 4 5 6 7 8 9 10 10 4/T f K l !--"

Fig. A-1 : Diffusion Coefficients of Fission Product Species in U02 as Function of Temperature

MODEL 123 Appendix A. TRANSPORT DATA FOR DIFFUSION n

ö c+1 M

N a

N Ln kn v v üN ä c~1 CV Q

Q Nn Itt "-+ 00nd' --" 00N W) Min NO~v)4 QIN tf) Q1 QN- tr) cr% 00 N --~ -+ N Ct' N --~ N u M -N

O L+ V] rr N~ r~rrtirr , a% m "" 00 \D GN 00 Ln 10 dl, 0O O~ M 1 "-" i~' ~rUe i 1 ir`r i t f r .~ ß+ MMMO~ Or~c+» ON~Q~ M ~ ~a \0 cV vri cV Ln c1l tn cV %,c - kn cq V-i ,--~

C UU UU U UU DDocUc ä° Q ccö°Ic M M cc I~ ~G C) .. ..~ .-, N d ,... ,_. cd c) Ö E~ QÖ ~ÖNQ 4Ö U Nö N~''N I° Q

"0.,

G . . ci

~ÜcnQx ~c3 UL,) ¢x ÜÜv~x " Qx a

124

'v=

~-- Temperature [`CI 2100 1800 1600 1400

4 5 6 7 104/T ( K1 ----y

Fig. A-2: Diffusion Coefficients of Fission Product Species in LTI Pyrocarbon as Function of Temperature

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL 125 a: N N NNtl)NNN O Ln 0C

O tf I~ N M 00 ON V'1 (~ u aN

N 'tN u tir \/ N ~ 00 ä Q U U U Q,. h C N W) W') \10 %.c tn tr) I- V1 l-- i-- ID It M U N O - l`bt-vnNN ~NN Q .NN r- N - 1~0 M \.D N LW.. u

aa v-, cn N N Ü ILI .'. 121 1-1 b O u vN tiG h 1-N \C N 00 l- %.c t- n ü O --* M ~D "D - c+t VI +-: 116 M 00 +--: N en w A

0UUU0 0 UI oU N UU 0 cz . 0 0 Ua ü 4 0D C w U C b M tr ~ O N V 0 C) CD C~ c U aä O N°° n V~ N O O ..r A U z r v

W Uc~Q

4.: c O n U V G C] G G C)

G O

ü O N ~

La L#*) c~ G O Ö O a m t) "~ Ü U a~ G s- n

~ v7 c~ 16 U 00 L . '- 00 U a U w :2 r-7 _ E . r- a> ° c U

ä ~-i L Q O C:

C v~. O 'U U G s U "'Z as cio _,c

0i ~'~! O O F c . .v~ V; .O G L-U E on X ~ G G

U va ¢ ^-c a ~m0

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL 127

10-4 -, FRG (1= Myers, upper curve for cesium ---- USA 2. Myers, lower curve for cesium) --- Japan 1u s -- Russian Fed.

Sic

O V

}- Temperature 1*01 107 20 2100 18001600 ' '~1200 1000

Fig . A-3a : Diffusion Coefficients of Fission Product Species in Silicon. Carbide as Function of Temperature

(a10- 13

Cesium in SiG

1()-14-, 1

1 reference 1986 -- reference 1991 1 auxiliary lines for fluence variation 10' 1 1

4- 4- 10F17-0

D l0-

18-16-19-,

F- Fast Neutron Fluence [1025 rr2, E>0 .1 MeV l

- Temperature VC 1 10- 20 2100 1800 1600, 1400 1200 1000 800 , 0 2 4 5 6 7 8 9 10 4 /T [K)

Fig. A-3b: KFA Diffusion Coefficients of Cesium in Silicon Carbide Comparison of 1986 and 1991 Reference Data and Fast Fluence Dependence

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL 129 h v; o i O N II p .A ü

a C~ q~ ü U as " M r- E II 0 U ,-, S- cz C)

cu Ö cM q ~ ö

0 0 0000 C v; 0 N M Cf' U 1+, u N ä

cz o 1-1 tot.. ' N R e-, 000 ~, r~ .r- ~ ~. u RT N ~o -r "t r- \W tiJ \r `1.r ü `. `r ~ " z" r.+ u ^~ ~o o1.coQ $ 00 tn11cr~ . II +' L. M ..~ N M vz oÖ M A g 0 .a" co

U U U U r aUUUUo a o ö ' M ~ O C7 bA ~ C» O Q ~n _' c a; A. o 0 o cV crs r w Et1~ NNN - 1 © Ü I t cu FU, VI VI VI VI n 00 o°` ~, a ö c ~ cet ~ ~ v "o üC> G Ü© U +, «S cr. öA ,~ U V L S. 011, o C) r3r U ''0 C 1--i 4-P cu u-, a vad-- ..~ ~ U~ cn xUcnd ~~ Z; ac' a C% o F rs, ~ ~ ~ p; ~Ä cn h ~

130

10-4

I (pores)

10s

-1a u. 10

12 y 1O

v- 0 0 c 10-14

C 10-1s Matrix

FRG , A3- 3 matrix graphite ---- Japan , matrix graphite for VHTR 10` 18 ---r United Kingdom, natural graphite matrix material ----Russian Fed. , matrix graphite - Temperature f V 10-20 2100 1800 1600 1404 1200 100o 800 4 5 6 7 8 9 10 10 4/T [K) ------"

Fig. A-4a: Diffusion Coefficients of Fission Product Species in Matrix Graphite as Function of Temperature

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL 13 1

10-4

Iodine in Matrix Graphite

10+6 -i

10~8

04 ,,,Xe (Canada) ,10 - 0 w C 10 03 C] 0 O O 0 O 0 v 10' 12 C 0 0 O 0 O

14 L1 1Ö - ISR Reference for graphite grains (after Müller) -- Xenon in U impregnated graphite o Optimized calculations of heating tests 10' Is -0---- Temperature 1°C] 2000 1800 1600 1400 1200 1000 4 5 6 7 8 104/ T ---~

Fig. A-4b: Diffusion Coefficients of Iodine and Fission Gases from Uranium Contamination in Graphite (Optimized Data Points Related to Contamination Fraction of 1 * 10'7)

Appendix A. TRA,'*,SI'©RT DATA FOR DIFFUSION MODEL 133 . (64 a en C' 'et kn M M \0 \,D t-- -r %-c \D IT lZ t~ t- t-- [-- r- t-

en `r

O N

u O N Q

U M

n (V

u N cCS F. O

. C ` a% 00 00 CIS 00 t` '"r N t d' et tn 03 ~' ~p~n0 'RT~ k+"~~,...~ n N~c O r+NN ---N -~cV+-- ~MM N 0 ä L

%D N \0 tiC N \G M ~t et N +-- ü~r %lr ~~ 12s~üv ü 'J t-~%C0 t- t-- 0 CO) N t- 00cn~c I ke~ oo .-: p C v7 UUU UU UUUU UUU U 0 0 O O O 0 0 O O 0 O O O .V +a G L70 00 0000 Ob 0 ~CQ v) 0 Mcr100 Q00 0 ~N00 0000 11C ~N U 0 00 CON 00 0000 00 0 c CD""' V') CD tn0 kn Ln00 ~LnWn o c. ~ 00 in 00 1- t- ~D %~G 0000 ON L w U .ä a

Q °n. U ~_ 0 2,-w UrnQ . ~0zn lu

13 4

1 (pares)

108 -

10 8 -

a E

4-0

0 14 ö 10 0 Structural Graphite

0 10-18 FRG " U$, H-451 structural graphite -- Japan , 1G-110 structural graphite --- United Kingdom, AGL-9 graphite 10-18 ~

-19 Temperature ('C 1 2100 1800 IB00 1400 1200 1000 800 10-2a

Product Species Fig. A-4c: Diffusion Coefficients of Fission Temperature in Structural Graphite as Function of

DIFFUSION MODEL 135 Appendix A. TRANSPORT DATA FOR

Cgf - Concentration [ mMole/kg l 10-s -

10'$ -

0 10-10 - N E

aD 0 0 10-14 ~! c Strontium

KFA , A3 - 3 matrix graphite -- GA , H- 451 structural graphite ---- UKAFA , AGL -9 graphite

- Temperature i'C l 2100 10- zo 1800 1600 1400 1200 1000 800

Fig. A-4d : Concentration Dependence of Diffusion Coefficient of Strontium in Graphite

10-4

10-6 C gr tt Concentration [ m Mole/Kg I

10`$ C gr =

a 100 1 10- io N \ \_ 10 r \ 100 .210- 12- 0 \ 10 2

14 _ Cesium ä 10

r- Temperature VC) 1000 800 2100 1800. 1600 1400 ,"1200 I I 4 5 6 7 8 9 10 10 4/T [K] --o--

Fig. A-4e: Concentration Dependence of Diffusion Coefficient of Cesium in Graphite

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL 13 7 c.-~ N M t+1 ~ d' cf' 'd' '1:31 d' d'

n Ö

u Q N ä öA w.

r-~ O G N O u cd N Ö "v Q ca

a o

v ~ o uC NO ä a .. 14

Ü C : LO Ö CT-

ü C:~ tr 'A M +-j O Ö "-+ CV N N N '0 N `r C U U U UU N öö 'C7 C (D C r oO NNN N 44 o Ö E C4 N 6 G U C U "O t.. ul w 0 t~ ~C. y w Q cr Im I-CU . cn4 U V] A7 9 ¢ aa a

138

10-6 -

Zr C 10-e i,

f USA 10 N 10 - Japan a -- Russian Fed. E -

C 10-16 -

10-18 -

thN-- Temperature I *C 1 800 10-20 2101_10 1800 1 16001 1 1400! 1 12001 1 10001 1 1- 1

Fig. A-5: Diffusion Coefficients of Fission Product Species in Zirconium Carbide as Function of Temperature

Appendix A. TRANSPORT DATA FOR DIFFUSION MODEL 139 140

Appendix B. INPUT DATA FOR PARTICLE FAILURE MODELS UNDER ACCIDENT CONDITIONS

The FRG particle failure model PANAMA consists of two failure mechanisms for TRISO particles under accident conditions. The coupling of these failure mechanisms, eventually extended by the manufacture-induced defect fraction 00, is then given by:

In the pressure vessel model dominant in the lower temperature range, the failure of a particle occurs as soon as the stress imposed upon the SiC layer by the internal gas pressure exceeds the tensile strength. The fraction of particles that fail due to pressure vessel at time t after the initiation of an accident at a temperature T is given by (Ref. 21):

= 1 -- e- in 2 " ( 47o }m (B _ (DI (t}T) 2)

where ct : stress induced in the SiC layer due to the internal gas pressure [Pa] co : SiC tensile strength at the end of irradiation [Pa] m : Weibull modulus

The strength values for SiC are scattered in accordance with a Weibull distribution where m specifies the Weibull parameter.

is determined based on the assumption of the SiC as a thin shell The stress at pressure vessel . Its thickness is further reduced by fission product corrosive attack on the silicon carbide. A "thinning rate" as function of temperature can be . given by Montgomery employing experimental data on SiC corrosion rates (Ref 22):

B . INPUT DATA FOR PARTICLE FAILURE MODELS UNDER ACCIDENT Appendix CONDITIONS 141

where kC : frequency factor for SiC corrosion [s -1] d0 : initial SiC layer thickness [m] QC : activation energy of SiC corrosion [.7/mol] = 179000 Qc R : gas constant, R = 8 .3143 [J/(mol K)] T : temperature [K]

The ideal gas law is taken to calculate the internal gas pressure depending on the geometry of kernel and void volume, the yield of stable fission gases, the gas release from the kernel (Booth formula), the heavy metal burnup, and the produced number of oxygen atoms per fission resulting in CO formation. The latter is strongly dependent on the particle type : it is highest for UO2 (Ref. 191), comparably low for (Th,U)O2 (Ref. 192) and zero for UCO (Ref. 193).

The Weibull modulus as well as the layer strength are considered as silicon carbide material properties and have been measured for various unirradiated particle batches (Ref. 23) . These values, however, are not available for the SiC used in the German reference particle batch EUO 2308. For safety analyses of German HTGR designs, corresponding data of the particle batch EO 1607 have been taken representing some sort of medium data (see table B-1) .

Further strength measurements of the batch EO 1607 after irradiation in the test HFR-GM1 (irradiation temperature TB - 1165 ° C, neutron dose r = 3 .7 1025 m-2, E> 0 .1 MeV) have revealed a fast neutron fluence induced decrease of the medium SiC strength and a broadening of its Weibull distribution (Ref. 21), i.e . a lower value for m leading to a higher failure probability of the pressure vessel . The functional dependence m ^ B and a = deduced m(r,T ) o a0(r,TB) from this single irradiation experiment is used in the PANAMA-1 code as a general rule. It thus may introduce an uncertainty into the calculation and may be a reason for the conservative results for failure fractions of particles irradiated to very high fast neutron fluences (Ref. 156).

The thermal decomposition dominant at very high temperatures beyond 2000 °C is represented by a parametrization which always remains S 1 :

142

e_a ~Q 02 (t,T) = I - '

where C a so-called "action integral" ~ = f kD (T) dt et, ß- empirical constants

The action integral C comprises the entire temperature-time history experienced by the particles. The parameters a and Q are different for loose particles or for particles in a fuel sphere and have been derived from temperature ramp tests up to 2500 °C of DR-S6 particles (Ref. 194) and of AVR-G02 spherical fuel elements (Ref. 120), respectively. Values are given in table B-l:

Thermal decomposition causes a measurable weight lass of the SiC layer which is again interpreted as a "thinning rate":

e-kd (T) " 3.75 " 102 . Qd 1(R " f) (B - 5)

where kd: frequency factor for SiC thermal decomposition [s -l] Qd: activation energy of SiC thermal, decomposition [Jfmoll Qd = 556000

exhibiting a significantly higher activation energy for the thermal decomposition process.

Predicted values for particle failure fractions under accident conditions of small modular HTGRs are expected to be very low even if an uncertainty factor as proposed in Ref. I is is taken into account.

results for accident 18 To account for the uncertainty of PANAMA-I model calculations, the by a factor of 20. With temperatures S 1600 °C were recommended in Ref. I to be multiplied to zero (i.e. a factor of I on increasing temperatures the uncertainty range should then decrease the results) when reaching 1800 °C.

INPUT DATA. FOR PARTICLE FAILURE MODELS UNDER ACCIDENT Appendix B. CONDITION'S 143 In the US, the Integrated Failure and Release Model for Standard Particles is used as reference model (Ref. 28). Its empirical approach comprises a SiC corrosion and a SiC thermal decomposition process which are, together with the SiC layer thickness, variables with statistical Weibull distributions. The basic equation reads as follows:

(D - - " C m )(t,1,21 e- (B - 6)

where : a so-called "action" integral t = f k(T) dt k : frequency factor for failure [s - k - k . * e-Qi/(R*T) o, L with i = c : Sic corrosion and i = d : SiC thermal decomposition R : gas constant, R = 8 .3143 [J/(mol K)] T : temperature [K] t : time [s] m : Weibull modulus

ko c : pre-exponential constant (corrosion) [s-1] log ko = - A * 3*log Ti + log c fsic A : constant Ti : irradiation temperature [K]

fSic' fission density [m-3] Qc : activation energy of Sic corrosion [J/moll

Qc = 252000 kold : pre-exponential constant (thermal decomposition) [s -1 ] log ko d = - 1 .58 + 2 .67*log T i + 0 .61*log I- r' : fast neutron fluence [10 25 /m2 , E > 0 .18 MeV]

Qd : activation energy of Sic thermal decomposition [J/mol] Qd = 556000

The two failure mechanisms are combined to a total failure fraction according to equation (B-1). Data are listed in table B-1.

The difference in the activation energies for SiC corrosion Qc between equations (B-3) and (B-6) is due to the fact that the 179.5 kJ/mol value is based upon all available data including the accelerated irradiation test data (Ref. 22). The 252

kJJmol is a revised number that comes from dropping the accelerated irradiation data which were supposed to be not representative for HTGR heating conditions (Ref. 195).

A revised version of this model based on an extended set of heating test data is considered to replace the original version as reference model (Ref. 8) . A more recent analysis of the activation energy by Goodin has found the observed temperature dependence of FRG heating tests to be essentially equal to the activation energy of SiC thermal decomposition. This observation was interpreted such that the thermal decomposition process was the only significant failure mechanism.

A pressure vessel failure is considered insignificant compared to the above degradation processes. However, data for a mean SiC strength of US particles and its Weibull distribution have been recommended to be used in model calculations (Ref. 8) (see table B-1) .

The Japanese particle failure model uses an approach similar to PANAMA-1 also based upon a thin-shell pressure vessel model. The failure fraction is determined by (Ref. 25):

where cr : stress induced in the SiC layer due to the internal gas pressure [Pa] vo : characteristic SiC strength [Pa] m : Weibull modulus

with the SiC strength obeying Weibull statistics. The strength vo is defined to be a function of the porosity P in the SiC layer introduced by thermal dissociation:

Appendix B. INPUT DATA FOR PARTICLE FAILURE MODELS UNDER ACCIDENT CONDITIONS 145

P co aoo ' e- n - [Pa]

and 4 eP A , t , - Q1(R " 7) (B - 9)

where doo : constant [Pa] n: constant A: constant [s -1] t: time [s] Q : activation energy [kJ/mot]

Data are given in table B-I.

Additional features of the JAERI model are the treatment of the statistical variation in the number of particles and the stoichiometry of the fuel kernel .

Good agreement has been found between model predictions and postirradiation heating test experimental results. Table B-1: Constants Used in Particle Failure Models

FRG SiC Layer Strength (EO 1607, unirradiated): 834. MPa Weibull Modulus (EO 1607, unirradiated): 8.02 Activation Energies corrosion Qc: 179.5 kJ/mol thermal decomposition Qd: 556. kJ/mol Constants Loose Particles: a: 0.693 ß: 0.88 Particles in Fuel Sphere: a: 0.693 ß: 0.88 USA SiC Layer Strength (after irradiation): 480. MPa Weibull Modulus (after irradiation): 5. Activation Energies corrosion Q.: 252. kJ/mol thermal decomposition Qd: 545. kJ/mol Constants A for oxidic fuel : 36.53 for carbidic fuel: 36.70 Weibull Modulus corrosion mC: 3.3 for single experiments 1 .6 for core predictions thermal decomposition md: 2.2 for single experiments 1 .5 for core predictions Japan SiC Layer Strength (unirradiated): 1650 MPa Weibull Modulus (unirradiated): 5.6 Activation Energy Q: 912. kJ/mol Constants A: 5* 109 s-1 n: 12.5

ACCIDENT Appendix B. INPUT DATA FOR PARTICLE FAILURE MODELS UNDER CONDITIONS 147 148 Appendix C. SORPTION ISOTHERMS OF FISSION PRODUCTS OVER GRAPHITIC SURFACES

At the boundary between a solid and a gas, transition of diffusing atoms occur due to sorption processes. In most practical cases, these two processes are so fast that a local equilibrium between the concentration of atoms adsorbed at the solid surface and the concentration of atoms in the neighboring layer of gas ("vapor pressure") may be assumed . This means that the phenomenological data of importance are given by an isotherm which relates these two equilibrium values. Usually the equilibrium vapor pressure in the gas phase is expressed as a function of the fractional coverage of the solid surface by adsorbed atoms . In practice, adsorption is mainly of importance in the case of porous solids which have a large internal surface area per unit weight. The vapor pressure is expressed as an exponential function of temperature and sorbate concentration.

Polymeric carbon, coked phenolic resin binder, has a particularly high sorption capacity for cesium and strontium at high temperatures. This can be attributed to its structure. Its density is low compared to that of graphite. The material has a turbustratic structure. The carbon layers are not parallel and not planar and they have many defects. This high sorption capacity of A3-3 matrix graphite containing coked ungraphitized phenolic resin binder is substantially higher than that of nuclear graphites like H-451, H-327, JG-110, BAR 675, P3JHAid, ASR-IRS, ASR-2RS, and ATR-2E, which contain a more highly graphitized binder component.

The high values obtained for the isosteric enthalpies of sorption of cesium and strontium by the A3-3 matrix graphite show that they are strongly bound by chemisorption . Thus, the sharp reduction of the partial pressures by sorption, the high sorptive capacity, and the large quantity of the A3-3 matrix graphite present in the core of an HTGR indicate a high potential of the A3-3 matrix for the retention of cesium and strontium during normal operation and in the case of a core heatup accident. This potential increases on irradiation with fast neutrons, if the irradiation temperature is below 1135 °C, and decreases if irradiation temperatures reach 1400 °C (Ref. 188).

At low concentrations, the sorption conforms to Henry's law (constant heat of adsorption, direct proportionality between vapor pressure and concentration of

Appendix C. SORPTION ISOTHERMS OF FISSION PRODUCTS OVER GRAPHITIC SURFACES 149

sorbed species) and at higher concentrations Freundlich sorption holds (characterized by decreasing heat of adsorption with increasing concentration of sorbed species) .

The fission product concentration sorbed on the carbonaceous material is in equilibrium with the partial vapor pressure, p, of that fission product species. The partial vapor pressure is assumed to contain contributions from both the Freundlich (PF) and Henrian (pH) isotherms (Ref. 165):

P _-_ PF + PH

In PF = (A + ) + ( D + -E, ) " In cgr (C -' 2)

In PH = (A + T ) + ( D -- 1 + T ) " In ci + In cg,, (C - 3)

d In cr = dl - 2 " T (C - 4)

where Freundlich isotherm vapor pressure [Pa] PF : pH : Henrian isotherm vapor pressure [Pa] T : Temperature [K] cgraphite : concentration of sorbate species [mmol/kg C] ct : transition concentration between the Freundlich and the Henrian sorption isotherm [mmol/kg C] A, D, d l : constants ß, E : constants [CC] d2 : constants [K -1]

Sorption isotherms of metallic fission products on different graphitic materials have been thoroughly measured in the past. At KFA, experiments on two dif ferent facilities have been carried out to derive sorption isotherms mostly for cesium and strontium on A3 matrix graphite (Ref. 188).

In the US design, sorption effects are important both for the porous fuel compacts and for the graphite blocks. The fuel compact consists of fuel particles and graphite shim particles, bonded by a sorptive carbonaceous matrix. The same governing equations are used for the sorption isotherms for fuel compact and H-451 structural graphite.

Design equations which incorporate the effect of fast fluence on metallic fission product sorptivity in H-451 graphite have also been derived as a modification of the equations (C-2) and (C-3) . These equations for cesium are reported in Ref. 177 and will not be repeated here. The sorptivity of the compact matrix material does not change with increasing fast neutron fluence (Ref. 164) .

The cesium concentration sorbed on compact matrix material tends to be one to two orders of magnitude higher than on unirradiated H-451 graphite; this difference narrows at higher pressures (10-6 Pa and higher) and lower temperatures . Cesium sorption on irradiated H-451 is closer to that of the compact matrix material. Compared to cesium, the difference in sorbate concentration between the two materials is greater for rubidium but less for strontium.

Values of the constants A, B, D, E for the sorption isotherms and d l , d2 for the transition concentration are given in Table C-I.

Figs. C-1 and C-2 show the sorption isotherms of cesium and strontium over matrix graphite as proposed by Moormann (Ref. 184). These data are preferred to be used in predictive calculations to those published by the experimenters Hiipert (Refs. 181 and 185) and Kwasny (Ref. 182) . The reason is that the latter describe the experimental results within the measured temperature and concentration ranges as given by the vertical lines in the figures. The new Moormann data set creates (almost) the same sorption isotherms in the Henry regime and overcomes the problem of "unphysical" behavior of the original equations in the Freundlich regime at higher temperatures beyond those covered by the experiments (Ref. 184) .

The sorption isotherms of cesium, strontium, and iodine on US structural graphite H-451 are plotted in Figs. C-3 through C-5.

Appendix C. SORPTION ISOTHERMS OF FISSION PRODUCTS OVER GRAPHITIC SURFACES 15 1

The data in Table C-ll indicate the method by which sorption isotherms are represented in the FRESCO model both for fuel element matrix graphite and for reflector graphite (Ref. 159) . Based on the ideal gas law, the partition coefficient of fission products between graphite and coolant is expressed by the ratio of graphitic surface layer concentrations in the coolant (cgas) and in the (cgraphite) or:

Cgas - a ' Cgraphite (C - 5)

dimensions for instance a is dimensionless; cgas and c:graphite have both the same "Atoms/m 3". For the linear range of the sorption isotherm is:

BH R + a . e H T (C _ 6) ffenry T

else

_BF ~F ) " e Ap + T -F ( DF + ~., " In cgrophire .,.... ']) aFreundlich -- T (C

where AH , BH : coefficients for the linear range of sorption isotherm (Henry) AF,BF,DF,EF : coefficients for the non-linear range of sorption isotherm (Freundlich)

graphite' concentration of sorbate species (The transformation factor of dimension /m3t, cgraphite "Atoms o£ equation (C-5) into the cgr dimension

if mmol/kg" of equation (C-7) is hidden in the above coefficients) .

The modeling of an unresisting transition between gaseous and solid phase is given by equating the partition coefficients with 1 .

The transformation of the constants A, B, D, E, as given in the equations (C-2) and (C-3) into the constants AH, BH, AF, BF, DF, EF, as given in the equations (C-d) and (C-7), is defined by the following equations:19

For the Henry sorption isotherms:

InAH -- A - ( - R) + (D - 1 ) - In cf 103

BH = B + E " In ct

For the Freundlich sorption isotherms:

InAF = A - ( p " R ) 103

BF '= B

DF = D - 1

EF = E

where p : graphite density, p = 1750 [kg/m3] for A3-Matrix R : gas constant, R = 8 .3143 [J/(mol K)]

19 Please note: these transformation equations are strongly dependent on the above given units for concentrations (here: mmol/kg) and pressures (here: Pa).

Appendix C. SORPTION ISOTHERMS OF FISSION PRODUCTS OVER GRAPHITIC SURFACES 153 Table C-II presents three data sets for cesium and strontium on A3-3 matrix graphite recommending uncertainty margins of the reference values for use in safety analysis calculations. They have been defined in Ref. 1 by factors of 2 (strontium) and 10 (cesium) on the partial pressure in the temperature range covered by the experiments and an additional 5 % on the sorption energies at higher temperatures. kn Nc}'--+Cf'O ~~Od'M 00 00 00 00 00 Q~ 1~0 110 01 % ~10 00

M 00000 1 ~O r cy .~..0 'L3

6- c'l NNI--I-- CMN I" NNNC%c~ 1 Q)M 'C ., oc x 0 - M . cli c-i cri ri 1 1 1

I" M M M 'd' A+ -' ~ 00 M00~00 /~ 0 --00 N 00 00 M W G N V 0t N C a

00`ci'~-rM 00 --In .c c% r- zr ~c N cr - N p ~Mt"11n0%C) +nMOlnln McVÖ~O .~G ..." ,.." oö 1 1 1 i 1 i 1 1

.h.l

r-l V" OMcr1Ni'-k y 0©In0Cl 1_-1 MCT00knN\0W nQ~N~ .~ Irr roq r~`° °`° n e°r`° d~-st 1 1 1 1 1 1 1 1 1 1 i v ta"n C 'Q' %n C% 0 cM 0 00 c+1 M ed (:Nt'-In00en OMMMM lrM Q _ as AG~d'0~Cf' ' N NNQN .4

C 0. y "~ ;~ yC >G 4G C1 4G r-. .-, cz '- 'c. 'w. 'i. 'C a ä 0 ä It b,~

r cnCMMMcnx ~xzL)U Ö yCy -C A A A re) A LL. EL ßly. c~ QQQQQ w+ Ü U ~ Ca., Z U U Ü In zn U" va~' ,-I U rr~ F ~ a

Appendix C. SORPTION ISOTHERMS OF FISSION PRODUCTS OVER GRAPHITIC SURFACES 155

N \Z it) O O N .N... ccs OO~c+1 c; u'1 11», v1~ncirr1 N 1 0,% CD 00 NMCN~ 9 r- O U Po CD :t M1 MF 1 1

lf~ r

I1 .r 1 . a oa N n ~ U U ö CD rriM ~ NN N ; oC C NInN- r nQ M C M Lri 1 1 n r 1 V3 V "0 N Q C L ca O s7, s., 00 c km = ~N ~' ..rm Q r N u w N NÖNä O n ~ n tp v'1 w cb 1 F ~ r r DiJ O V V ü M 1 O (Z rn u d r~ cd G et äß 0 w 'ZJ -r 00 M rr1 N v'1 cd O Ö00 C? O e 0 N N ÖN ^~ ^ ~" NQN~O'~i 1 A,. U 1 M ~ ~ Q CC 1 ~ t 1 " ctS w a ü w

G q U ~" ro U U x j 0 c U

QWQmmw~ [ QmQcnLlw

F 0

Cesium on matrix graphite

l0-4 ld-2 10 ° l0 2 Concentration [ mMole l kg ] ---~

Fig. C-1 : Sorption Isotherms (Reference Data) of Cesium on A3-3 Matrix Graphite

Appendix C. SORPTION ISOTHERMS OF FISSION PRODUCTS OVER GRAPHITIC SURFACES 157

Strontium on matrix graphite

T in °C

10-a a

10- 15 -

_a ä 10 Measure- cti ments KFA

10-12 _ Uncertainty 10f I / margin 10 14 1-~-I - I i 10-4 10 z 10" 10 2 Concentration [ mMole I kg 1--"

Fig. C-2: Sorption Isotherms (Reference Data) of Strontium on A3-3 Matrix Graphite

- I T - i - -1 10-4 10- 3 10-2 10- ' 10° 10' Concentration [ mMoie 1 kg ] --o-

Fig. C-3 : Sorption Isotherms (Reference Data) of Cesium on H-451 Structural Graphite

Appendix C. SORPTION ISOTHERMS OF FISSION PRODUCTS OVER GRAPHITIC SURFACES 159

10-4 Jö 3 Id-2 10-' 160 161 Concentration [ mMole 1 kg I -.

Fig. C-4: Sorption Isotherms (Reference Data) of Strontium on H-451 Structural Graphite

Iodine on H-451 graphite

Tin°G

10F, -

16-7 10 6 10-5 10-4 10-3 10 2 101 Concentration [ mMole / kg ] -----0-

Fig. C-5: Sorption Isotherms (Reference Data) of Iodine on H-451 Structural Graphite

Appendix C. SORPTION ISOTHERMS OF FISSION PRODUCTS OVER GRAPHITIC SURFACES 161 FORSCHUNGSZENTRUM JULICH GmbH

JUI-2721 January 1993 ISSN 0366-0885