The preceding part of the chapter has been written with breeder react­ors in mind. Different fuel is needed for a reactor designed to consume plutonium rather than breed it, the principal difference being that it should not contain uranium or at least that it should have a minimum of it. There is little irradiation experience with high-Pu fuels so in what follows it is possible only to indicate what seems most likely to be of use in practice.

If plutonium is to be consumed then it is clearly desirable that it should be consumed as quickly as possible. Plutonium is consumed by fissioning it, which produces energy. The power of a plutonium­consuming reactor should therefore be high if it is to be effective. A 2500 MW (heat) reactor with an 80% load factor, fuelled with pure plutonium with no breeding, would consume about 800 kg of plutonium per year.

The constraints of heat transport (see section 3.2) mean that a 2500 MW core would have to have a volume of about 3 m3 and would need some 1700 kg of plutonium to make it critical. If the plutonium were in fuel elements generating an average of 32 kWm-1 (i. e. a max­imum rating of 50 kWm-1 at the centre of the core) the fuel elements would have to have a total length of 78 km. If the fuel were PuO2 with an 80% smear density the radius of the pellets would be about 1 mm, and if it were metal or some other ceramic the radius would be even smaller! Clearly this is impossibly small for practical fuel elements: they have to be larger, and in practice a diameter of about 6 mm is the minimum that is economically and structurally possible. Thus the plutonium has to be diluted with some inert material, which in effect is present to replace the uranium that makes up such a large part of the fuel of a breeder reactor.

There are two ways in which a diluent material can be incorporated in a ceramic fuel. It can either be a solid solution in the ceramic, or a second component in a two-component mixture. Because of the difficulty of dissolving PuO2 in nitric acid for Purex reprocessing, and also because it reacts chemically with sodium, plutonium oxide solid solutions such as (Pu, Zr)O2 are not attractive. The most promising ceramic solid solution is (Pu, Zr)N. There are few data on its physical properties but theoretical estimates suggest that, as in the case of Pu-Zr alloy, it would perform well in a 50 kWm-1 fuel element.

In the 1960s, when a solution to the problem of fuel swelling was being sought, “cermet” fuels were investigated. A cermet is a sintered mixture of ceramic and metal powders. The idea was that the metal would form a strong matrix that would constrain the swelling of the fuel ceramic. Some irradiation testing was done and high burnups of UO2-stainless steel cermets were achieved. However, the project was abandoned when it was realised that if the metal fraction was high enough to prevent swelling it would absorb so many neutrons that the breeding ratio would be reduced unacceptably. This of course would not be a disadvantage in a plutonium-consuming reactor, and cermets of PuO2 with steel or other metals such as chromium, vanadium or tungsten may be attractive.

By analogy with a cermet, a sintered mixture of two mutually insol­uble ceramic powders is sometimes called a “cercer”. The range of suit­able diluent ceramics is restricted because the resulting cercer would have to be soluble in nitric acid if the fuel is to be reprocessed by the Purex process. Cercers of PuO2 and magnesia (MgO) or yttria (Y2O3) are possibilities. Ceria (CeO2) should probably be ruled out because it is incompatible with liquid sodium and would swell severely in the event of a cladding failure.

Because of the success of U-Pu-Zr alloy fuel Pu-Zr alloy appears attractive. The melting point of plutonium is very low at 640 °C but that of zirconium is 1852 °C. The solidus temperatures of Pu-Zr containing 20% and 40% Pu are about 1680 °C and 1480 °C respectively. Since their thermal conductivities (without porosity) are around 20 Wm-1 K-1, comparison with the 70U-20Pu-10Zr fuel of section 2.5.1 seems to indicate that a high-Pu metal-fuelled element would have acceptable irradiation performance at a linear rating of 50 kWm-1.

Figure 4.13 shows the temperatures of the coolants of a 3600 MW (heat) reactor with once-through steam generators. Steam is supplied to the turbine at 18.5 MPa and 490 °C at a rate of 1650 kg s-1. The final feed temperature is 240 °C. The total secondary sodium flow rate is 15330 kg s-1 and the primary sodium flow rate is 18935 kg s-1.

The choices available in designing the steam generators can be illustrated in a simplified way as follows. Suppose there are N tubes each of length L and diameter D. Then we have

Q ~ UATmNnDL, (4.4)

where ATm is the logarithmic mean temperature difference, Q is the total heat transfer rate, and U is the mean heat transfer coefficient which depends on the steam-side conditions. If the total mass flow­rate in the evaporators is M, then clearly M = mN.

A once-through steam generator can be thought of as consisting of three regions as far as heat transfer is concerned: the inlet region where the tube is filled with single-phase water, the boiling region with a two-phase mixture of water and steam, and the superheating region of steam in the vapour phase. The heat transfer coefficient in the superheating region is considerably lower than those in the regions where liquid water is present. Values for the heat transfer coefficients for the three regions are given for example by Collier (1972), and can be used to find U, which depends on m and therefore on N, but not very strongly.

D has to be at least 15-20 mm for ease of manufacture. For a 20 mm tube, m = 1kg s-1 makes the mean speed of the two-phase mixture in the centre of the steam generator about 12 m s. If m is much greater than this the problems of vibration and erosion are significant. If D is taken to be about 20 mm the product LN is limited by equation 4.4 and the main choice is between large L (long tubes) or large N (many tubes).

Figure 4.14 shows the sodium, steam/water and tube mid-wall tem­peratures in a steam generator of the reactor described in Figure 4.13, with D = 20 mm and m = 1kgs-1. This choice of m implies that there would be 1650 tubes each about 64 m long. Clearly this rules out straight tubes, but it is reasonable for reasonably compact bundles

of helical tubes. An alternative design might select larger diameter tubes. D = 28 mm would give m = 0.5 kg s-1 and N = 3300. In this case, allowing for the slightly lower values of U, the tubes would be 30 m long and could possibly be arranged in six separate straight-tube steam generator units each with 550 tubes.

The discontinuities in the mid-wall temperature shown in Figure 4.14 mark the beginning and end of the two-phase region. Because of the turbulence caused by the formation of vapour heat transfer in this region is better than in the pure liquid region and much better than in the vapour region. An important consequence is a marked step in the tube temperature at the “dry-out” point — the last point where the surface of the tube is wetted. The dry-out point is not fixed but moves up and down the tube both as operating conditions change and at random due to the turbulent nature of the two-phase flow. Thermal stresses are associated with the temperature step, and as they fluctuate there is a possibility of fatigue damage to the tube material. (The extent of these stresses is limited, however, because whatever the heat transfer conditions the tube temperature is constrained to lie between the sodium and steam temperatures that, because of the good heat transfer on the sodium side, are relatively close. In a fossil-fuelled plant where the tube would be heated by a flame at a very much higher temperature the potential is for wider stress fluctuations and therefore greater fatigue damage.)

Equation 1.1 is far too complicated to be solved analytically without drastic simplification. Before it can be solved numerically it has to be recast with the continuous independent variables changed into discrete forms. This can be done in various ways, as follows. To make the explanation simpler we shall assume we are dealing with a reactor that is operating steadily so that we can ignore t.

Position. The position vector r is discretised in the form of a spa­tial mesh covering the reactor core and the surrounding breeder or consumer regions. It is usually convenient to make the geometry of the mesh coincide as far as possible with the actual structure of the core. As explained in Chapters 2 and 3, most fast reactor cores consist of hexagonal fuel subassemblies so a fully three-dimensional mesh is usually either hexagonal or triangular in the radial directions and linear in the axial. (This is in contrast to the square calculation meshes for thermal reactors, the cores of which are usually made up of square subassemblies.) For some purposes a cylindrically symmetric two-dimensional (r, z) approximation, linear in both directions, may be adequate.

Having defined the spatial mesh equation 1.2 can be recast in terms of the mesh intervals. There are two ways to do this. In the finite difference method ф is calculated at each mesh point, whereas in the nodal method the average value of ф over the volume of each cell of the mesh is calculated.

Energy. E is discretised by dividing the range of energy of interest (from about 10 MeV down to 0.1 eV or lower) into a series of intervals Eo — Ei, Ei — E2, etc., where E0 may be 10 MeV, and Ei < E0, E2 < Ei, etc. The neutrons with energies between Eg and Eg-1 are known as group-g neutrons. The flux фg in group g is then given by

Eg+1

ф^г, ft) = ф(г, E, ft)dE. (1.5)

Eg

If there are G groups £r, £f and v are vectors with G elements. £s becomes a G x G matrix the nature of which depends on how the directional variables ft are discretised.

Direction. The direction vector ft can be discretised in either of two ways. One is to specify a set of points on the surface of an imaginary unit sphere and then recast equation 1.1 in terms of the flux of neutrons moving in the directions of areas dft surrounding these points. This leads to “Sn ” methods, where n is the number of discrete directions. Alternatively ф can be expressed as the sum of an infinite series of spherical harmonic components, and equation 1.1 rewritten in terms of the coefficients of the series. In the case of a one-dimensional axially symmetric approximation

TO

ф = ^2 a«P« (cos в ), (1.6)

n=0

where the Pn are Legendre polynomials. This is called a “Pn” method. In either Sn or Pn calculations the greater the number of “mesh points” (i. e. the number of directions for Sn or terms of the series for Pn) the greater the accuracy. In practice n = 5 is usually found adequate.

At low temperature heat is conducted mainly by the diffusion of phon­ons through the crystals. As the temperature increases the density of phonons increases and, as the behaviour of the crystal lattice is slightly
nonlinear, the probability of interaction between phonons increases, their mean-free-path decreases and the conductivity decreases. At higher temperature still electrons become more mobile and make a growing contribution to the conductivity.

For mixed oxide the conductivity is affected by the chemical com­position, which can be described as (PuaU(1-a))O2+x, where a rep­resents the plutonium fraction and x, which can be either positive or negative, the departure from the stoichiometric oxygen content. It is usually in the range -0.03 < x <0 .03.

The effect of stoichiometry on conductivity is as follows. If there is no plutonium (a = 0), below about 1400 °C Kf decreases steadily as x increases, but above 1600 °C Kf is a maximum for x = 0 and is reduced by departure from the stoichiometric composition in either direction. This seems to be because at high temperature an excess or a deficiency of oxygen, not exceeding a few per cent, is accommodated by vacancies or interstitials in the lattice. These defects reduce the phonon mean — free-path and hence the conductivity. At low temperatures, however, excess uranium (x < 0) is precipitated as the metal and the resulting free electrons increase the conductivity.

The situation is quite different if plutonium is present. Uranium and plutonium have nearly the same ionic radius and stoichiomet­ric UO2 and PuO2 form solid solutions in whatever ratio they are mixed (for all values of a). But if the mixed oxide is not stoichiometric (x < 0) the excess metal is accommodated by means of lattice defects for x > -0.02 and by precipitation of Pu2O3 for x < -0.02. Thus for mixed oxide with a Pu fraction typical of a large fast reactor (a ~ 0.2) the conductivity decreases as the composition departs from stoi­chiometric in either direction (x < 0 and x > 0) at all temperatures. The effect is quite large: for x = +/-0.02, Kf is reduced to about 75% of its value for x = 0. For x = 0 the conductivity decreases slowly as a increases, so that in the range 400-1200 °C, for a = 0.2 Kf is about 13% smaller than for a = 0.

This picture is confused in practice by structural effects that can be much more important than the effects of composition. Principal among these is the effect of porosity, which reduces the effective conductivity. As explained later (section 2.3.2) the fuel may be manufactured with 10% or more porosity (i. e. 10% of the overall volume of the fuel may be occupied by voids). Porosity of 10% can reduce the thermal effective conductivity by as much as 25%.

Estimation of the effect of porosity in an operating fuel element is made very difficult because, as explained in section 2.4.1, the pores move under the influence of the temperature distribution, so that the conductivity changes with time and in different ways in different parts of the fuel. Further confusion is provided by the cracks in the fuel that are formed under the influence of differential thermal expansion and that have a large but unpredictable effect on conduction. Finally fission products, which themselves move through the fuel (section 2.4.6), and changes in crystal structure (section 2.4.1) affect conductivity in a manner that is not accurately known.

The maximum linear heat rating q is set by the requirement that the fuel should not melt. The melting points of UO2 and PuO2 are about 2850 °C and 2430 °C respectively, and the solidus temperature for a = 0.2 is about 2730 °C. Figure 2.1 shows the variation of Kf (T) and f Kf (T)dT with T for unirradiated stoichiometric (U0.8Pu0.2)O2 from 500 °C to the melting point. The integral from a fuel surface temperature Tfs of 1000 °C up to the melting point is about 4.9 kWm-1 for stoichiometric oxide.

Equation 2.3 shows that f Kf (T)dT — 4.9 kWm-1 implies q ~ 62 kWm-1 if the fuel has no central hole. But if there is a hole with radius 20% of the fuel radius (a = .04) equation 2.5 shows that q can be increased to about 71 kWm-1. In practice, because of the uncertainty of the effects of cracking, porosity, stoichiometry, changes in composition with burnup and the conductance between fuel and cladding (section 2.2.3), q is limited to about 50 kWm-1.

Under the influence of irradiation materials creep at low temperatures. The precise mechanisms of irradiation creep are not certain but in general terms it is clear that, as atoms displaced from their sites move about, they rearrange the crystal structure in such a way as to reduce the elastic energy, and if the material is under stress this gives rise to strain in the direction of the stress. For example one mechanism by which this might happen is that interstitial atoms produced by neutron scattering may tend to migrate to defects such as edge dislocations, causing them to climb so that the material strains. Another possible mechanism is that vacancies may coalesce on a plane in the crystal and if there is a compressive stress normal to the plane the resulting disc-shaped void may collapse causing the material to strain. At high temperature thermal agitation gives rise to such creep mechanisms: irradiation allows creep to take place at much lower temperatures. For example, substantial irradiation creep has been observed in 316 stainless steel at 280 °C whereas thermal creep is not significant below about 600 °C.

The creep strain is in most cases proportional to the stress, almost proportional to D, and nearly independent of temperature. Figure 3.14

shows the dependence of the ratio of creep shear strain to stress on D for various materials.

Irradiation both hardens materials (i. e. it raises the yield stress) and makes them more brittle (i. e. it reduces the elongation before failure). The ultimate stress usually changes relatively little, but there is a loss of work hardening. Typical stress-strain curves are shown in Figure 3.15.

There are two mechanisms that cause these effects. At low test temperatures the defects caused by irradiation damage reduce the mobility of dislocations and inhibit plastic strain so that the uniform elongation is very low. At higher test temperatures (500-600 °C in 316 stainless steel), the material anneals, the defects are removed, and the properties of the unirradiated material tend to be recovered.

Above 700 °C in 316 stainless steel the second mechanism comes into play. As shown in Figure 3.15 there is a loss of ductility which can outweigh the increase in ductility in the unirradiated material due to thermal effects. This is the result of helium generated by (n, a) reactions. At high temperatures it diffuses to the grain boundaries

where it collects in the form of small bubbles. These cause a loss of cohesion between the grains, so that they can be torn apart.

Radioactive materials have to be kept from being released into the environment both in normal operation of the reactor and in the after­math of an accident. There are usually three substantial containment boundaries: the cladding of the fuel elements, the primary coolant

containment (i. e. the reactor vessel in a pool layout, or the reactor, pump and heat exchanger vessels and the connecting pipework in a loop layout), and the reactor building. Fuel and fission products are contained by all three; activation products including 24Na by the last two.

Failure of the fuel cladding is discussed in section 2.4.7. The fuel is designed so that failure of any one element is very unlikely, but there are so many elements in a reactor core (of the order of 105 in a 3000 MW (heat) reactor, each being renewed after each year or so of operation at full power) that the possibility of some failures has to be allowed for. Failure by fission-product corrosion or due to a defect in manufacture is likely to result in no more than a small crack in the cladding, which would release to the coolant just the fission-product gases and possibly some volatile fission products. The amount of radioactivity reaching the coolant would be small unless a much larger breach was made in the first place, or the small breach was enlarged by operating the reactor for a long time without replacing the failed fuel. Experiments have shown that even in the event of gross cladding failure the amount of fuel released is very small (Smith et al., 1978).

If there should be widespread cladding failure gaseous fission products would find their way to the cover gas over the coolant. They would be contained by the roof of the reactor vessel, but there might be some leakage through the seals on pump shafts, control rod actu­ators and rotating shields. The seals have to be designed to keep this, together with the leakage of 24Na, to a low level.

Apart from such minor leaks, the primary coolant containment also has to be designed to prevent a major breach. This means that it has to be able to withstand the effects of accidental loads that might be imposed from without, by a piece of heavy equipment being dropped, for example, or from within by a core accident (see section 5.4.4).

Any release of radioactive material from the primary coolant cir­cuit is contained by the reactor building, or “secondary containment” as it is often known. The building has to have a ventilation system with suitable traps and filters to control any radioactivity released to the atmosphere inside the building and to cope with the effects of the sodium fire that would result from a breach of the primary or second­ary coolant circuits. (Ways of preventing such a fire are described later in section 5.3.1.) The building itself has to withstand loads imposed by wind or snow, by earthquakes (if the reactor is to be built in a zone where these occur), by missiles from external sources such as an explo­sion in a nearby piece of plant or equipment, or by a crashing aircraft.

In summary although care has to be taken in design it is reasonably straightforward to ensure that the three containment boundaries are independent and will remain intact under the loads imposed by a wide range of less severe accidents, which are the ones most likely to occur. This is the principal foundation of the safety of a reactor of this type. Once this foundation has been established release of radioactivity is possible only if there are coincident independent failures of all three boundaries, which is very unlikely, or if some single initiating event can cause breaches in all three.

If the core had the same composition throughout the power density would be nearly proportional to the flux and would be distributed across the core as shown on the left of Figure 1.12. This would be most undesirable from a thermodynamic point of view because if the coolant flowed at the same rate through all parts of the core it would emerge at different temperatures. When cold coolant from the periphery of the core mixed with hot coolant from the centre there would be a gain of entropy and a consequent loss of work output, and there would be large temperature fluctuations in the mixing region that might damage the structure (see Chapter 3). Alternatively if the flow in the outer part of the core were restricted to equalise the outlet temperatures pumping work would be wasted as the coolant flowed through the restrictions. For this reason the core is normally made in two or more radial zones, with higher fuel enrichment in the outer zones. The effect of this is to increase the power density in the outer region as shown on

the right of Figure 1.12, thus reducing the coolant outlet temperature differences.

Figure 1.13 shows the radial distribution of neutron flux and power density in a small breeder reactor having two core zones, of roughly equal volume, with enrichments of 22% and 28%, surrounded by a breeder. If the core were uniform the enrichment would be about 24%. The peak power densities at the centre of the core and the inside of the outer zone are roughly the same and the radial power peaking factor, Pmax/Pave, where Pmax is the power generated in the most highly rated channel and Pave is the average power per channel, is reduced from about 1.35 in a single-zone core to 1.21 in the two-zone core.

An important point to notice in Figure 1.13 is the change in the power density at the inside of the breeder with time. As fissile material is generated in the breeder, predominantly at its inside edge, it in turn undergoes fission and generates power. Over the life of the inner fuel elements of the radial breeder (which would stay in the reactor much longer than the core fuel) the power density rises considerably. In this particular reactor at the start of its life the power density at the centre of an inner radial breeder fuel element is 60 MW m-3 when the reactor power is operating at 600 MW (heat), but this rises to 220 MW m-3 after 1.6 years at power.

The fuel swells at a roughly constant rate throughout irradiation and has usually closed the gap and come into contact with the cladding after about 0.1% burnup. The cladding also swells (section 3.3.2) but only slowly at first, more rapidly later, and at a rate that depends strongly on temperature. The result is that for some cladding materials and in some parts of the core the swelling rate of the cladding may eventually exceed that of the fuel, tending to reduce the stress between the two and reopen the gap.

While cladding and fuel are in contact and there is a compressive stress between them as shown in Figure 2.12 both creep. Swelling of the central part of the fuel inside the load-bearing ring is accommod­ated by expansion into the central void. Swelling of the outer part is accommodated partly by the swelling of the cladding.

The cladding strain is not always uniform and in some cases may be concentrated by cracks in the fuel pellets, and this stress concentration, rather than uniform strain, may determine the maximum burnup to which the fuel can be subjected. Sometimes distortion of the pellets can cause non-uniform strain of the cladding. During irradiation a pellet tends to distort into the shape shown, very much exaggerated, in Figure 2.13. The ends tend to be displaced outwards as shown and can

cause ridges round the cladding, giving it the appearance of a bamboo cane. These ridges are not usually permanent because the pellets soon fuse together, but they are sometimes observed in fuel that has been subjected to abnormal conditions.

Some early fast reactors used electromagnetic pumps to circulate the coolant, which have the advantage that no moving part penetrates the sodium containment. The sodium is pumped either by passing an electric current through it in the presence of a transverse magnetic field (a conduction pump) or by subjecting it to a moving magnetic field (an induction pump). However it proved difficult to scale elec­tromagnetic pumps up to the size needed for large reactors and now mechanical pumps are used universally. The problem of penetrating the sodium containment can be met by means of electric motors with totally enclosed, “canned”, rotors, but the usual method is to allow the shaft to pass through the containment above the sodium level. The penetration is thus exposed to the argon cover gas containing sodium vapour but not to liquid sodium, and oil seals have been found satisfactory in most cases.

The pumps are thus driven by motors situated on the roof of the reactor vessel (in the case of the primary pumps of a pool reactor) or the pump vessel (in the cases of a loop reactor and the secondary circuits of either style), via vertical shafts with thrust bearings at the top and sodium-lubricated sleeve bearings at the bottom. The long shafts of the primary pumps of a pool reactor are usually tubular with a large diameter to avoid whirling. The variable-speed motors are fitted with auxiliary or “pony” motors that are capable of turning the pump fast enough to maintain adequate flow to keep the fuel cool when the reactor is shut down. In the case of power failure these can be energised from a standby source such as a diesel generator to guarantee emergency cooling.

The pumps have to provide a large volume rate of flow at a relat­ively low pressure rise. A typical 3600 MW(heat) reactor would have three primary pumps each delivering about 8 m3 s-1 at 500 kPa. Single­stage centrifugal impellers of conventional design are normally used. In most cases the primary pumps deliver the coolant vertically down­wards and the pump volutes are designed so that the resulting axial thrust, which can be as much as 10 tonnes, is borne by the pump casing rather than the shaft.

The main difficulties in design are to cope with sudden changes in temperature and to prevent cavitation, and in these respects loop and pool reactors pose different demands. The coolant temperature has very little effect on cavitation, because even at 600 °C the saturation pressure is only 7 kPa. The important factor is the pressure at the pump inlet (the “net positive suction head”). The pressure of the cover gas above the sodium in the reactor or pump vessels is limited to some 1-200 kPa gauge to minimise leakage of radioactive material and sodium vapour. The pump inlet pressure is this gas pressure plus the hydrostatic head of the sodium above the pump. The risk of cavitation can be minimised by increasing the depth of immersion, decreasing the rotational speed (which implies increasing the rotor diameter) or shrouding the inlet to make the flow uniform. For the primary pumps of a pool reactor all these options imply increases in the dimensions of the primary vessel.

In a loop reactor, however, the suction pressure depends on whether the pump is located before or after the heat exchanger (i. e. whether it is in the “hot leg” or the “cold leg”). If it is before the heat exchanger there is a loss of pressure due to the flow in the pipe from the reactor vessel to the pump. If it is after the heat exchanger there is an additional loss of pressure due to the flow through the heat exchanger and more pipe. It is not easy to compensate for these pressure drops by positioning the pump at a lower level to increase the hydrostatic head, and anyway doing so would require still longer pipes. Thus the conclusion is usually reached that cavitation is avoided more easily if the pump is placed in the hot leg, before the heat exchanger. The relative advantages of hot leg and cold leg pumps are discussed by Campbell (1973).

Primary or secondary pumps may be exposed to sudden temperat­ure changes in the event of emergency shutdowns (“trips”) of either the reactor or the steam plant. This is not usually a problem in the case of the primary pumps of a pool reactor because they draw from the large mass (of the order of 2000 tonnes or more) of cold coolant filling the vessel. In a loop reactor, however, the situation is quite dif­ferent. If the reactor is tripped a hot leg pump is subject to a rapid fall in temperature, and if a secondary heat exchanger is shut down because of a steam plant trip a cold leg pump is subject to a rapid rise. These temperature transients can be designed for, but they constitute a disadvantage of the loop layout.

The history of fast breeder reactors is quite dissimilar from that of thermal reactors. From the earliest days after the Second World War the development of different types of thermal reactor was pur­sued in different countries: light-water reactors in the United States, heavy-water reactors in Canada and gas-cooled reactors in the United

Kingdom, for example. Only towards the end of the 20th century did the various nationally based lines of development converge.

In contrast virtually the same path was followed in all the countries where work on fast reactors was done. The reason for this seems to have been that until the 1960s at least fast reactors were seen to be commercially valuable only well into the future, so that the advant­ages of cooperation appeared to outweigh the disadvantages of aiding possible competitors. Thermal reactors on the other hand were com­mercially important from the start and were developed in competition, which restricted the exchange of ideas and allowed different concepts to flourish.

International cooperation played a major role in fast reactor devel­opment for two main reasons. Firstly the nuclear data on which designs had to be based were inadequate until the 1960s. There was a lot to be gained from worldwide cooperation in measuring neutron cross­sections to the required accuracy and exchanging and comparing the results. Secondly cooperation to ensure the safety of fast reactors was desirable even when there was competition in other areas.

This need to exchange information resulted in, among other things, a series of international conferences on fast reactors that were addressed mainly to the problems of reactor physics and safety and were held in the United States and various European countries throughout the 1960s and 1970s. These, together with the continual publication of information in the scientific press, kept the think­ing in different countries from diverging and encouraged parallel development.

In one respect it is not altogether certain that this was an advant­age. The use of liquid metals as coolants acquired a great deal of momentum, mainly because “everyone did it”, and the search for alternatives was discouraged. Gas has certain advantages as a coolant, but at the time of writing no group has been able to develop a gas — cooled fast reactor to the point where it can be assessed fairly in comparison with a liquid-metal-cooled reactor.