Before about 1960 it was thought that a high breeding ratio was the most important quality of a fast reactor. To achieve this the mean energy of the neutrons has to be kept high, and this requires that extraneous materials, especially moderators, should be excluded as far as possible from the reactor core. As a result the early reactors had metal fuel, the metal being either enriched uranium or plutonium, alloyed in some cases with molybdenum to stabilise it to allow opera­tion at higher temperature.

The critical masses of these reactors were small and the cores were correspondingly small so that for high-power operation they had to be cooled by a high-density coolant (to avoid impossibly high coolant velocities). Hydrogenous substances were precluded because hydro­gen is a moderator, so liquid metals were used. In most cases the coolant was sodium or sodium-potassium alloy. Some early experi­mental reactors were cooled with mercury but this fell out of favour because of its toxicity, cost, and low boiling point.

The many neutrons that leaked from the small cores of these reactors were absorbed in surrounding regions of natural or depleted uranium where the majority of the breeding took place. These were known as breeders, or blankets.

The first generation of low-power experimental fast reactors were built in the late 1940s and early 1950s to demonstrate the principle of breeding and to obtain nuclear data. They included CLEMENTINE and EBR-I in the United States, BR-1 and BR-2 in the Union of Soviet Socialist Republics, and ZEPHYR and ZEUS in the United King­dom. CLEMENTINE, ZEPHYR, and BR-1 and 2 used plutonium fuel, which in the early years was more readily available than highly enriched uranium. Apart from ZEUS, which was a zero-power mock — up of the later DFR, they all had very small cores, the largest being EBR-I (6 litres), which was a small power reactor with an output of 1.2 MW.

When it came to higher powers, however, the volume of the core had to be increased to keep the heat fluxes down to reasonable levels and to allow for the extra coolant flow. The result was EBR-II and EFFBR (the Enrico Fermi Fast Breeder Reactor) in the United States, and DFR (the Dounreay Fast Reactor) in the United Kingdom. When they were designed they were seen as prototypes of the reactors to be used in power stations, but as they were built it began to be recognised that they would be the end-point of the development of metal-fuelled fast reactors, and the principal use to which DFR and, for many years, EBR-II were put was to test oxide fuel for the next generation.

In normal operation the temperatures of the various materials in the core vary in a regular way and it is useful to sum up their combined effects. There are various ways of doing this, one of which is to determ­ine the “isothermal temperature coefficient”. This is the rate of change of reactivity with temperature if the temperatures throughout the core change by the same amount. It includes the effects of axial and radial expansion of the structure, expansion of the coolant, and the Doppler effect. Because of the variation of the Doppler coefficient it depends slightly on the fuel temperatures. For a typical sodium-cooled breeder it might be -2.5 x 10-5 K-1 when the fuel is cold (i. e. the reactor is operating at very low power), and -1.5 x 10-5 K-1 at normal oper­ating temperature. It can be thought of as measuring the response of the reactor to a change in coolant inlet temperature when power and coolant flow-rate are held constant.

Another useful parameter is the power coefficient. This is the rate of change of reactivity with power, assuming the coolant flow-rate changes proportionately to the power and the inlet temperature is constant. The coolant and structure temperatures everywhere are thus constant so that the power coefficient depends only on the Doppler effect. A typical value of the power coefficient for a 2500 MW (heat) sodium-cooled breeder is -3 x l0-6 (MW)-1.

The most stringent requirements are those for selecting fuel for a ded­icated consumer reactor fuelled with minor actinides alone, mainly because the linear rating of the fuel has to be as high as in a conven­tional reactor. Both the thermal conductivity and the melting point are lower for minor actinide oxides than for UO2 or PuO2. This rules out single-component oxide fuel. Similarly the thermal conductivity of americium-bearing oxide cercers is so low that realistic linear ratings cannot be achieved. The situation is better for oxide cermets as far as linear rating is concerned, but unless the metal fraction in the cermet is to be kept around 50 v/o or more it cannot be expected to seal the fuel, so in the event of cladding failure there is a risk of swelling on contact with Na. Thus oxide in any form is effectively ruled out for a dedicated

sodium-cooled consumer, although oxide cermets might be possible in consumers cooled by lead or gas. The remaining possibilities are

• Pure compound (Am, Cm)N

• Solid solutions (Am, Cm, Zr)N or (An, Cm, Y)N

• Cercers (Am, Cm)N+TiN or (Am, Cm)N+AlN

• Cermets (Am, Cm)N+W

An alternative to a reactor fuelled entirely with minor actinides is a “normal” plutonium-fuelled reactor with special fuel subassemblies containing the minor actinides in or around the core. The power rating of these “target” assemblies would not have to be so high, especially if

they were placed round the periphery of the core, so they could make use of oxide fuel.

Pure minor actinide oxides are ruled out because they are unlikely to be stable under irradiation, but it may be possible to stabilise them in solid solution with Zr or Y to give a fuel suitable for targets. Similarly the low thermal conductivity of Am-bearing oxide cercers may be acceptable. The lower rating of the targets might also make cladding failure sufficiently unlikely that swelling of the fuel material on contact with sodium is less important, allowing oxide cermets or cercers to be considered. The possibilities for target fuel include all those identified above and in addition the following:

• Solid solutions (Am, Cm, Zr)O2 or (Am, Cm, Y)2O3

• Cercers (Am, Cm)O2+MgO

• Cermets (Am, Cm)O2+W, Cr, V or steel

The use of the secondary sodium circuit to separate steam from radio­active primary sodium carries the disadvantage of a loss of thermal efficiency due to the increase of entropy as heat is transferred from primary to secondary sodium. The extent of the loss can be estimated very conveniently in terms of the available energy (sometimes called the exergy). The specific steady-flow available energy a, is given by

a — h — T0s,

where h and s are specific enthalpy and entropy respectively, and T0 is the dead-state or environmental temperature.

If the working fluid in a cyclic power plant circulates at a rate M and receives heat as it changes from state 1 to state 2, the rate at which it receives heat is

Q — M(h — h2)

and the maximum work output Pmax is given by

^max — M(a1 a2 )•

The maximum thermal efficiency is Pmax/Q. Pmax is attained only if all the processes taking place in the plant are reversible. Equation 4.7 shows that Pmax is the rate of increase of available energy of the work­ing fluid as it receives heat. Every time heat is transferred — from the fuel to the primary coolant, from primary to secondary coolant and from secondary to water and steam — enthalpy is conserved but entropy is increased and therefore the available energy is decreased (equation 4.5).

Table 4.1 shows the steady increase of entropy of the fuel, the primary coolant, the secondary and the steam in a 3600 MW (heat) plant. The entropy of the fuel is greater than that of the source of the energy in the fission fragments, because their kinetic energy has been turned into disorganised thermal agitation (i. e. into heat). The other increases in entropy are due to the transfers of heat from higher to lower temperatures.

Table 4.2 shows the lost potential for doing work represented by each increase in entropy (assuming an environmental temperature of 300 °K). The increase in available energy of the steam in the steam generator is 1860 MW, and this would be the net power output if the steam cycle were reversible. There are, however, various irreversibilit­ies — in the turbine for example, due to pressure losses in pipes, and in the feed train — so that the actual net work output would be 1490 MW.

The overall efficiency of the plant is 41%. It is of interest to note that the greatest sources of loss of thermal efficiency — i. e. the greatest

sources of entropy production — are the change of the kinetic energy of the fission fragments into heat in the fuel, and the transfer of that heat from the fuel to the coolant. In contrast the loss of efficiency due to the presence of the secondary sodium circuit is a mere 0.2%, which is small compared with the losses due to irreversibility within the steam plant. The flow of available energy is shown diagrammatically in Figure 4.16.

After integration over all directions, some manipulation, and the discretisation of energy into G groups, the steady-state version of equation 1.1 for a homogeneous region of space gives G linked equa­tions, one for each of the group fluxes fg,

g-1 1 G

0 = DgV2fg — Vrgfg + Esg^gfg + kXg Vg Efgfg+ Sg. (1.9)

g’=1 g’=1

Equations 1.9 are the multigroup diffusion equations. The terms on the right-hand side of equations 1.1 and 1.9 correspond to each other. The groups are numbered in reverse order of energy, so that group

1 contains the neutrons of highest energy and group G those of the lowest. Thus the limits of summation in the third term of equation 1.9 correspond to the absence of up-scattering. The diffusion coefficients Dg are given by Dg = 1/3Esg(1 — ~fig) where д is the mean cosine of the angle of scattering in the laboratory system of coordinates. The boundary conditions at an interface between regions with different properties are that фЕ and Jg = Dg grad фЕ should be continuous.

For a reactor driven by a source (Sg ф 0) к is made equal to 1 and the фg represent the resulting flux distribution. If no solution is possible the reactor assembly is supercritical. For a critical reactor with Sg = 0 к can be thought of in two ways. The straightforward interpretation is that it is the effective multiplication constant, keff, of the reactor. But if к ф 1 the reactor cannot be steady and the фg cannot be constant, so that under this interpretation equations 1.9 have a meaning for к = 1 only. Alternatively к can be thought of as the highest eigenvalue of the set of equations that can be solved to find к and the ф^ although in this case the фg have no physical meaning unless к = 1. The composition of the reactor has to be altered to make к = 1, and when this is achieved the фg are the group fluxes in the critical reactor.

If the neutrons are treated in only one group — that is, if differences of energy can be ignored — the third term in equation 1.9 disappears and it becomes

0 = DV2ф + (vVf /к — Va )p + S. (1.10)

J2a = J2c + f has replaced r because neutrons are removed only by capture or fission.

Before the multigroup equations can be solved of course values have to be given to the constants Dg, Vrg, Vsg^g, xgvg and Vfg that depend on the microscopic cross-sections for the various materials in the reactor. Because the cross-sections vary with neutron energy the group constants involve average values over the energy range covered by each group.

In principle it is possible to make the groups so narrow that the variation of each cross-section within each group is small. However, in practice this cannot be done because the very complex fine structure of the variation of the cross-sections with energy, especially in the resonance region, would necessitate an impossibly large number of
groups. A method has to be found for calculating suitable group cross­sections to facilitate accurate calculations with 20 or 30 groups.

Almost every fission event turns one atom (of uranium or plutonium) into two (of fission products). Some of these are gaseous or volatile (see section 2.3.5) and are mostly lost from the fuel, but the majority are solid and are retained either as interstitial atoms in the crystals or as inclusions of a distinct solid phase. Some of the gaseous fission products are retained as small bubbles within the crystals or on the grain boundaries.

Very roughly all atoms occupy about the same volume (of the order of 10-29 m3) in solid and liquid phases so the increase in the number of atoms means that the volume increases. Loss of some fission products from the fuel tends to mitigate the increase, but on the other hand retention of bubbles of gaseous fission products tends to enhance it. The net result is a fractional increase in the volume of the solid fuel of about 0.8 times the burnup.

This swelling is almost inexorable and if the fuel fits closely inside the cladding when it is new the cladding is forced to strain to accom­modate it. Thus for example 10% burnup would imply about 3% normal strain in the cladding in both the circumferential and (because friction between fuel and cladding does not allow relative motion) axial directions. As pointed out in section 3.3.4 the ductility of the irradiated cladding may be too low to accommodate so large a strain. Since the primary purpose of the cladding is to retain the radioactive fission products, if more than a very few fuel elements crack because of the swelling of the fuel inside them the design is unacceptable. The burnup has to be limited by design to a value at which trial irradiation has shown that no more than an acceptable number of cladding failures will occur.

It is usual to allow for the effects of fuel swelling by leaving space for the fuel to swell without straining the cladding. In some cases this has been done by manufacturing the fuel in the form of pellets with a central hole so that the increase in volume can be accommodated by filling or partially filling the hole. Alternatively and more usually the fuel is manufactured as porous pellets, the pores occupying some 10-20% of the overall volume, and the swelling is accommodated by filling the pores.

It is by no means certain, however, that either pores or a cent­ral void work in the manner intended to accommodate swelling. As explained in section 2.4.1 pores migrate during operation and a central void is often formed even when one is not present initially. In addition the fuel creeps (see section 2.4.3) so that there is a tendency for stress between fuel and cladding to be relieved by strain of the fuel as well as of the cladding, and irradiation-induced reduction of the density of the cladding (section 3.3.2) also helps to accommodate fuel swelling. In fact the picture is so confused that, although the strain of the clad­ding of an irradiated fuel element can be measured, it is not easy to predict.

A convenient parameter often used to characterise the space avail­able to accommodate the swelling of the fuel is the “smear density”. It is often defined as the ratio between the mass of fuel per unit length of the fuel element as manufactured and the mass of fuel it would contain if the cladding were completely filled with fuel at its maximum theoretical density (i. e. if there were no gap between the pellets and the cladding, and no porosity or central hole in the pellets). Oxide fuel elements are typically manufactured with 80% smear density.

Sometimes the term is used slightly differently, being defined as the average density the fuel would have if it were spread uniformly across the cross-section of the fuel element. Thus if the theoretical density of UO2 is 11000 kgm-3, a fuel element with 80% smear density according to the definition in the previous paragraph might alternatively be said to have a smear density of 8800 kgm-3.

Iron, chromium and especially nickel are soluble in liquid lead. The solubilities are shown in Figure 3.18. In lead-bismuth eutectic (44.5Pb, 55.5Bi) the solubilities are about a factor of 10 higher. Unprotec­ted steel surfaces corrode very rapidly by dissolution, not only of the metal surface but also beneath the surface as the liquid penetrates into defects and along grain boundaries.

The corrosion rate can be reduced substantially by a protective oxide layer on the surface of the steel. At temperatures in the range of 300-470 °C and oxygen concentrations of 10-9-10-8 by atoms (~10-8- 10-7% by weight) an oxide layer a few tens of qm thick is formed that protects the steel from direct contact with the liquid metal. Corrosion continues by oxidation as oxygen diffuses through the layer. The layer grows in thickness until an equilibrium is reached, with oxide being formed at its inner surface and being lost by spalling from the outer surface. Figure 3.19 shows the composition of the oxide layer formed

on a ferritic steel after exposure to flowing lead-bismuth eutectic at 470 °C. The outer part of the layer consists of porous Fe3O4 whereas the inner part is a compact (Fe, Cr)3O4 spinel. The protection is only partially effective, however, in that corrosion by oxidation continues at rates of 50 |Fm per year or more. Similar but thinner protective layers are formed on austenitic steels.

The steel surface is protected only while the oxide layer is in place. It can be disrupted mechanically by the impact of particles carried in the coolant, or even by the turbulence of the coolant itself. The pressure fluctuations in turbulent fluid flowing with velocity v are of the order of pv2/2, and the high density of lead and lead-bismuth eutectic (for both of which p ~ 10 500 kgm-3) implies that the velocity has to be limited if the oxide layer is not to be eroded. In practice coolant velocities have to be kept below about 3 ms-1 to control the rate of corrosion.

Temperatures have to be limited as well. Above about 470 °C oxide is laid down in an increasingly thicker but less compact and more unstable layer that is much more susceptible to erosion, and above 550 °C it offers little protection so that both austenitic and ferritic steels corrode rapidly by dissolution.

There is some evidence that the presence of additional materials can result in the formation of more stable protective layer. A few percent of silicon or aluminium in the composition of the steel, for example, are beneficial in this respect. Alternatively it may be possible to aluminise the surface of the steel. Oxide-disperse steels (see section 3.3.8) may also be resistant to corrosion at higher temperatures. But although it is possible that such materials may provide a solution to high-temperature corrosion they require testing over extensive periods before they can be validated for use in a power reactor. Until this has been done corrosion effectively limits the maximum temperature in a reactor cooled by lead or led-bismuth eutectic to 470 °C.

A defect in a steam tube or a weld in a steam generator of a sodium-cooled reactor may give rise to a leak in the form of a jet of

500

0.25 0.5 1 2 4

Moles of sodium per mole of steam

Figure 5.5 Sodium-water reaction temperatures.

high-pressure water or steam into the secondary sodium. The water and sodium react chemically to produce sodium hydroxide, which is strongly alkaline, and hydrogen. The reaction between sodium and liquid water is

2 Na + 2 H2O ^ 2 NaOH + H2 + 8 MJ per kg of water.

Figure 5.5 indicates the temperature of the reaction products as a function of the ratio of the reactants, assuming the reaction is adiabatic and the reactants start at 350 °C with the water as liquid. If the reactants are in the stoichiometric ratio the products reach a temperature of about 1450 °C in steady flow.

Formation of sodium hydroxide is not the whole story because in practice some sodium oxide is formed by the reaction

2 Na + H2O ^ Na2O + H2 + 7 MJ per kg of water,

Self wastage

Original leak

Figure 5.6 Propagation of a small steam generator leak.

but this makes little difference to the temperature of the reaction products.

Small Steam-Generator Leaks. A small hole, such as a crack in a weld or a defect in a steam tube, might be equivalent to a circular hole of diameter 0.1-0.2 mm and give rise to a leak of steam or water into sodium at a rate of 10-100 mg s-1. This would form a hot under­sodium “flame” that could be very damaging. If the “flame” were to play on an adjacent tube it would heat and soften it, the caustic reaction products would corrode it, and the supersonic steam flow would erode the corroded steel and “blow it away”, until the tube was penetrated. The water or steam flowing in the tube would act to cool it and delay penetration for tens of seconds or longer depending on the size of the original leak, but eventually the tube would fail and the leak would propagate. Figure 5.6 illustrates the processes taking place as a small leak propagates in this way.

The steam generator has to be equipped with a protective system to minimise the damage caused by a small leak; the system has to prevent the leak from propagating and also prevent the spread of caustic reaction products that might damage the secondary sodium
circuit and in particular the intermediate heat exchangers. The system acts to detect the leak, and then to take protective action.

A small leak can be detected by monitoring pressure, the presence of hydrogen, or acoustic noise. Excess pressure in the expansion tank is normally used to actuate a trip system. It is reliable as a means of detecting a leak but not particularly sensitive in a large steam gen­erator. A hydrogen-in-sodium signal (as described earlier in section 4.2.7) can be used to actuate a trip but it may be a poor indicator of a leak because there is always some hydrogen present by diffusion through the steam generator tubes as oxidation takes place on the steam side. If the trip level is set too low spurious trips are unaccept­ably frequent. Hydrogen leak detection has the additional disadvant­age that it is difficult to design a system to respond quickly, in less than 10 s or so. Acoustic leak detection is attractive in principle but in practice it is difficult to discriminate against the noise of the coolant flow and the mechanical rattling of steam tubes against their support grids. Hans and Dumm (1977) survey in considerable detail the various methods of detecting leaks.

On detection of a small leak a trip is initiated to isolate the steam generator on both the steam and sodium sides. The steam is dumped through the safety valves and the intermediate heat exchanger is isol­ated to prevent contamination with sodium-water reaction products. Operation of the dump system trips both the reactor and the turbine. The isolation and dump system is shown diagrammatically in Figure 5.7. (The bursting discs would not operate in the event of a small leak.)

Once a leak is detected it has to be located and repaired. Acoustic methods may be useful for location, because if one side of the unit is pressurised it may be possible to hear the gas issuing from the leak. An alternative method is to seek sodium hydroxide on the water side of the tubes by chemical means, because it is found that sodium migrates through small leaks against the pressure difference. When the leak has been located the usual method of repair is to plug the affected tube or tubes.

Bursting disc

Effluent separator

Steam dump valve

Steam isolating valve

SDV Sodium dump valve

SDT Sodium dump tank

SIV

Figure 5.7 Steam generator isolation and dump systems.

In a steam generator made of austenitic steel there is a danger of more extensive damage because of the susceptibility to caustic stress- corrosion cracking. If after manufacture parts of the unit such as the tubeplate welds are left in a state of stress, and there is a leak nearby, the sodium hydroxide formed may cause cracking of the stressed region (Broomfield and Smedley, 1979). This is an important dis­advantage of austenitic steels in steam generators and is one of the reasons for the use of ferritic materials.

Large Steam-Generator Leaks. If there is a large leak, such as might be caused by a steam tube breaking in two (an event often known as a “double-ended guillotine failure” or DEGF), water or steam would be ejected into the sodium at a rate of the order of 1 kgs-1. Unless protective action is taken such an event might propagate to failure of other tubes. As in the case of a small leak a reaction “flame” would be formed, but it would be large enough to embrace several tubes. Somewhere in the flame region the reacting mixture would be in the

stoichiometric ratio above 1300 °C, hot enough to cause the steel of pressurised steam tubes to become soft and to burst. Figure 5.8 illus­trates the possible situation.

One kilogram of steam reacting with sodium generates about 1.5 m3 of hydrogen at 1 atmosphere and 350 °C, so in the event of a large leak (1 kg/s or more) large volumes of reaction products are formed very quickly. The secondary sodium circuit is exposed to the steam pressure of 16-20 MPa, possibly higher as the sodium and water react. The protective system has to relieve the pressure quickly. This is usually done by means of bursting discs that release the reaction products to an effluent system that traps the sodium and caustic material in a dump tank and vents the hydrogen and steam to atmosphere. Figure 5.9 shows such a system in outline.

When the bursting discs have opened the secondary sodium circuit is open to the atmosphere, so the intermediate heat exchangers are the only barrier between the radioactive primary sodium and the environ­ment. It is therefore essential that the integrity of the heat exchangers

should not be compromised. The capacity of the effluent system has to be sufficient to protect them from damaging pressures, and the isol­ation valves have to act quickly enough to prevent contamination by corrosive reaction products. Because of the possibility of propagation it may be necessary to size the bursting discs and effluent system to cope with the simultaneous severance of several steam tubes.

The most familiar application of fast reactors is as breeders, designed to turn fertile material into fissile. A simple performance indicator for a breeder reactor is the Breeding Ratio B, defined by

B = (rate of production of fissile material)/

where ^cfertiie is the macroscopic cross-section for neutron capture in fertile material and £afissile for absorption in fissile. The sums run over all energy groups and the integrals over the entire reactor including the breeder region.

B is widely used as a measure of the effectiveness of a breeder reactor, but it suffers from the major disadvantage that it does not take account of the differences between the isotopes. It counts an atom of 241Pu as being equivalent to an atom of 239Pu, whereas in fact the fission cross-section of 241Pu is higher and its capture cross-section is lower, so it is more valuable as a reactor fuel. A better measure of breeding is the “Breeding Gain” introduced by Baker and Ross (1963). This is based on an assessment of the values of the various isotopes as contributors to reactivity using perturbation theory in a simplified form involving a single energy group.

If there is only one neutron energy group we can see from equa­tions 1.27 and 1.30 that neutron flux and importance, ф and ф*, are proportional. If equation 1.25 is then rewritten for a single group we have

Sp a 5 (v £ f — £ f — £ c ф2 + SD (Уф)2 dV (1.42)

R

because £r = £f + £c, and the scattering terms are no longer signi­ficant. The macroscopic cross-sections are averages over the whole energy range weighted with фф*. Thus, for the fission cross-section,

£ f = £ Ntaft, (1.43)

І

where the subscript i represents the nuclides of which the reactor is composed, Ni is the number of nuclei of i per unit volume, and a fi is the average microscopic fission cross-section of i. The average is taken over the entire neutron energy range, weighted with фф*; i. e.

п TO I г — TO

a fi = a fi(E )ф(Е )ф* (E )dE ф(Е )ф* (E )dE, (1.44)

where ф (E) and ф* (E) are the fundamental mode flux and importance.

It follows from equation 1.42 that, if the effect of scattering is neglected, the reactivity increase when an atom of i is created is pro­portional to wi, where

Wi = va fi — a fi — a d. (1.45)

wi is the “worth” of an atom of i and measures its usefulness for building a new reactor of the same design. The “Breeding Gain”, G, is then defined as the net increase in worth (summed over all the nuclides present in the reactor) divided by the worth of the nuclides destroyed. It is still rather artificial because it assumes that the changes are distributed throughout the reactor in proportion to the numbers of nuclides present. (Thus fissile nuclides generated outside the core in a breeder region are assigned the worth they would have if they were in the core.) It nevertheless indicates the rate at which new cores, of the same specification, could be assembled when all the core and breeder fuel is reprocessed.

It is clear from the way they are defined that G & B — 1, but there is no algebraic way to show this to be the case.

Table 2.2 lists the fission products present in greatest abundance after 10% burnup of typical fast reactor fuel (30% plutonium with typical concentrations of the higher plutonium isotopes). The behaviour of a chemical system with so many components is obviously extremely complex and is certainly not understood in detail. The broad outlines are given in this section but the complexities are such that the actual behaviour in a particular fuel element with a slightly different com­position, cladding, temperature or irradiation history may differ quite widely. It is convenient to divide the most abundant fission products into groups as follows. Elements in the same group behave roughly similarly.

Inert Gases (Kr, Xe). These are mainly released from the fuel, but some are retained in solution or in small bubbles within the grains in the cooler parts of the fuel (see section 2.3.4).

Alkali Metals (Rb, Cs). These are very volatile in elemental form and migrate to the cool periphery of the fuel. This migration is illus­trated in Figure 2.16 which shows the distribution of 137Cs as determ­ined by y — spectroscopy. In some cases the isotopes that are daughters of inert gases, such as 133Cs which is produced from 133Xe which decays with a half-life of 5.3 d, and 87Rb produced from 87Kr decaying with a half-life of 78 min, can appear in the gas plenum to which the precurs­ors have migrated.

1.0

Figure 2.16 The distribution of fission products in irradiated fuel.

Metals forming refractory oxides (Sr, Y, Zr, Ba, La, Ce, Pr, Nd, Pm, Sm). By and large, having formed oxides these elements do not migrate and are found uniformly distributed through the fuel, as illustrated by the data for 144Ce in Figure 2.16. But again the isotopes that are daughters of volatile or gaseous precursors, such as 138Ba and 140Ba (from 138Cs and 140Cs respectively), are less uniform. 89Sr (a daughter of 89Rb that has a half-life of 15.4 min) migrates farther than 90Sr (from 90Rb, half-life 2.7 min).

Metals that do not form oxides (Tc, Ru, Rh, Pd, Ag, Te). These are found as metallic inclusions, sometimes dispersed through the fuel and sometimes, especially if the central temperature is high, having migrated to the central void. There they form droplets of metal that are molten while the reactor is operating and are found as small ingots when the fuel is examined subsequently. Such ingots usually contain uranium and plutonium as well. Figure 2.16 shows the loss of ruthenium from the hottest part of the fuel.

Metal that may form an oxide (Mo). The fate of the molybdenum depends on the oxygen potential of the fuel as explained in section 2.3.3. If it is low molybdenum is found in the metallic inclusions; if it is high it is found as MoO2.