The physics of a fast reactor depends on the materials that compose the core and on its size and shape but not strongly on the details of its structure, as explained in section 1.1.2. In broad terms the composition is determined as follows. The only material in the core that is essential to the physics is the fuel (unlike a thermal reactor in which the moder­ator is also essential), but the demands of heat transfer usually require that about 50% of the core volume is occupied by coolant. Structural material takes up another 20% or so leaving about 30% for fuel.

For a sodium-cooled power reactor in which the heat rating of the fuel is maximised the various demands of heat transfer and heat transport, which are discussed in Chapter 3, limit the height of the core to about 1 m and the average power density to about 500 MW m-3. The diameter of the core is then determined by the required power output, and criticality is adjusted by changing the proportion of fissile material in the fuel. For gas-cooled or lead-cooled reactors the power density is lower, possible about 200 MW m-3 or less, and the optimum core height is greater, possibly 1.5 m for gas coolant or 2 m for lead.

Most fast reactors that have been built so far, apart from experi­mental or test facilities, have been cooled with sodium and designed primarily to produce power.

2.1 INTRODUCTION

In common with pressurised water reactors (PWRs), boiling water reactors (BWRs) and advanced gas-cooled reactors (AGRs) most fast power reactors use oxide fuel. There is a certain amount of experience with metal fuel in the United States, and there is interest in the use of metal fuel and other ceramic fuels such as nitride or carbide for future fast reactors. This chapter deals mainly with the oxide and metal fuels of which there is most experience, and it covers other fuel materials in less detail.

Since oxide fuels are widely used much of this chapter applies as well to thermal reactors as to fast reactors. The main difference is that new fast reactor fuel usually consists of a mixture of plutonium and uranium dioxides whereas at present most thermal reactors use enriched uranium dioxide with a small amount of plutonium present after irradiation. The use of mixed uranium and plutonium diox­ide (MOX) fuel in thermal reactors is increasing. Most water-cooled thermal reactors have fuel clad in zirconium alloy, but AGRs use stainless steel, and because of the similarity of the coolant temperat­ures there is much in common between the behaviour of AGR fuel and that of a sodium-cooled fast reactor.

There is an important difference, however. In all systems it is eco­nomically desirable to irradiate the fuel as long as possible before it

is removed from the reactor. In a thermal reactor the limit is set by the loss of reactivity as fission products that absorb thermal neutrons accumulate. In a fast reactor the loss of reactivity is much smaller, especially if the internal breeding is high, so the irradiation limit is different. It is set by the need to be sure that the cladding will remain intact and not allow radioactive material to escape into the coolant. As a result the maximum burnup is very high in a fast reactor. The neutron irradiation accompanying this burnup damages the cladding and the irradiation limit is set by the need to remove the fuel before the cladding loses its integrity.

A very thorough account of reactor fuel in general, and in partic­ular of oxide fuel for fast reactors, is given by Olander (1976). It is particularly valuable for its treatment of the theory of many aspects of fuel behaviour.

3.3.1 Displacement of Atoms

The neutron irradiation to which the materials in the reactor are sub­ject alters their properties in several ways. The most important effects on the structural materials are to increase hardness and decrease ductility, to enhance creep rates at low temperatures, and, most import­ant, to reduce the density. These phenomena have to be taken into account in design of the reactor along with familiar effects such as thermal creep, fatigue and corrosion.

Irradiation affects the properties of non-fissile materials in two ways. Neutron scattering interactions displace atoms from their sites in the crystal lattice, creating vacancies and interstitial atoms in equal numbers, and neutron absorption by (n, a) and (n, p) interactions cre­ates atoms of helium, hydrogen and other transmutation products within the crystals. Helium has the greatest effect on the properties of the material.

A useful way to characterise the extent of the irradiation received by a piece of material is to specify the “displacement dose”, which is the average number of times an atom has been displaced from its lattice site. Each elastic scattering interaction imparts kinetic energy Ep to the target nucleus, where Ep is a random variable distributed uniformly (if the scattering is isotropic) in the range 0 — p. En, where En is the neutron energy, д = 4A/(1 + A)2, and A is the atomic weight of the target. For iron д = 0.069, so a 1 MeV neutron can impart up to 69 keV to an iron nucleus from which it is scattered elastically.

It requires only some 25 eV to displace an iron atom from its site in the crystal lattice, so the target nucleus may well have enough energy to displace several hundred atoms as it moves through the crystal. Some of the kinetic energy is however taken up by inelastic scattering interactions, and the number of atoms displaced by elastic scattering depends on the direction in which the target atom travels relative to the crystal lattice. Moreover some of the displaced atoms recombine with vacancies in the lattice. Various estimates of the relationship between Ep and nd, the number of atoms displaced, have been made. Figure 3.9 shows one such, that of Torrens, Robinson and Norgett (often called the “TRN”, or alternatively “NRT”, model), which is widely used.

It is possible to define a “displacement cross-section” for neutrons in group g, adg, by

r Eg-1 r Eg-1

odg ф(Е )dE = ae (E )n(E )ф(Е )dE, (3.12)

Eg Eg

where

(3.13)

ae(E) is the elastic scattering cross-section, and ф(Е) gives the vari­ation of flux within the group from, for example, a fundamental mode calculation. Figure 3.9 shows values of adg for iron.

The total number of displacements suffered by each atom, D, is then given by

(3.14)

where T is the length of time for which the irradiation continues. D is the average number of times each atom is displaced from its lattice site and is often referred to as “dpa” (displacements per atom).

The scattering cross-sections of chromium and nickel are quite similar to that of iron so it is often assumed that the displacement cross-section of iron as shown in Figure 3.9 can be used for all the constituent elements of the steel structural materials used in reactor cores, whatever their actual specifications.

Figure 3.9 shows that adg falls off rapidly at energies below about 0.1 MeV. For this reason an alternative way of characterising the extent
of irradiation, often used because it is easier to calculate, is the fast neutron fluence, Фf, defined by

(3.15)

The total flux above 0.1 Mev is sometimes called the “damage flux” and Фf is the “damage fluence”. If the power density Q in mixed-oxide fuel containing 20% plutonium is 2.5 GW m-2 (see section 3.2.1) and the neutron spectrum is similar to that shown in Figure 1.7 the damage flux is about 5 x 1019m-2s-1. If the fuel is irradiated to 20% burnup is about 1.4 x 1027m-2 and D is about 140 dpa. This irradiation takes about 5.4 x 107 s or 20 months at full power. These conditions are typical of the centre of the core and represent the full extent of the irradiation suffered by any of the structural material that is removed and replaced along with the fuel (i. e. cladding, subassembly wrappers, etc.). Any material in the core that is not replaced along with the fuel, such as the control rod guides, suffers a higher fluence. It is usual to design the core so that all of its structure can be replaced as necessary.

Failed Fuel. It is more difficult to detect an incipient accident affecting only one subassembly than one affecting the whole core, but it is necessary to do so to prevent the spread of damage (see section 5.4.2). The difficulty arises because there are hundreds of subassemblies in the core. If instruments have to be attached to each, and triplicated, the resulting trip system is very complex. It may also be very cumbersome, especially if instruments have to be fixed to each subassembly after it has been inserted into the core and detached before it is removed.

Thus instruments outside the core that can detect a developing accident anywhere in the core are very desirable. There are two main candidates: failed cladding detectors and boiling detectors.

A failed cladding detection system searches for failed fuel elements by monitoring the coolant or cover gas for radioactive fission products that must have come recently from the fuel, and then locates the subassembly in which the failure has occurred. This can be done in two ways. p or у activity can be sought if the fission products are separated from the coolant, which is already у-active. Alternatively neutrons from delayed neutron precursors can be sought if a sample of the coolant is removed from the neutron flux.

A typical system has four parts. A sample of the main coolant flow from the core is piped away to a point at which the neutron flux is low. There it passes to a vessel surrounded by moderating material and thermal neutron detectors (usually boron trifluoride counters) that detect any delayed neutron precursors in the coolant. This enables a quick response to cladding failures anywhere in the core, provided the sample has been taken from a point where the flow is well mixed so that it is representative of the flow through the whole core.

Secondly samples are taken from the outlet of each subassembly in turn to a delayed neutron monitor. This serves to locate the sub­assembly in which a failure has occurred. Because there are so many subassemblies to sample, location is relatively slow.

Thirdly a sample of the cover gas over the sodium in the reactor ves­sel is taken to a y-spectrometer and a moving-wire p — precipitator. This latter device uses a charged wire to precipitate the daughter products of p-decaying fission products and to transport them to a chamber con­taining a p-detector. The system thus discriminates against a range of activation products that might be present and selects p-active daugh­ters of gaseous p-active fission products (mainly 88Rb produced by p-decay of 88Kr), so that it responds selectively to cladding failures. Further information about the nuclides present is given by the y — spectrometer.

Finally the gaseous fission products are stripped from the loca­tion system coolant sample by a stream of gas that then goes to a p-precipitator. This provides an independent way of locating failed cladding.

Coolant Boiling. Failed fuel detection has the advantages that it is reliable and a direct indication of the release of radioactivity, which is what has to be avoided. Even a bulk detection system is however relatively slow, taking tens of seconds to detect a failed fuel element, mainly because of the time taken to transport the coolant sample to the detector. It is adequate to control the release of radioactivity to the coolant but a faster system can help to minimise damage to the fuel.

If fuel element failure is caused by overheating it may be accompan­ied by boiling of the coolant (which takes place at about 920-940 °C at the pressure in the core). A boiling detector gives a quicker indication that something is amiss than waiting for the fuel to fail and then for the failed fuel detection system to operate. Boiling can be detected by acoustic means, as is suggested by the ease with which boiling is heard in a domestic kettle. It is particularly attractive in a reactor because a small number of detectors are enough to detect boiling anywhere in the core.

The main difficulty is that the reactor itself is quite noisy. Sound is generated by the coolant pumps, by the turbulence of the coolant flow, and by cavitation. Cavitation is particularly awkward because it is a form of boiling (caused by local reductions in pressure at points where the flow of the coolant is accelerated, such as at sharp corners or on the blades of the pump impellers), and it makes a very similar noise to boiling caused by overheating. It may be possible to avoid the difficulty by designing the reactor to keep cavitation to a minimum and to discriminate against other background noises by listening in a frequency range in which boiling generates a lot of noise.

As an alternative to acoustic means boiling can be detected by temperature measurement. In some cases boiling in a subassembly can be detected by means of a thermocouple at the outlet, but there are some circumstances in which detection would not be reliable. A partial blockage to the flow somewhere in the subassembly could be large enough to cause severe overheating in its wake, possibly to the boiling point, but at the same time have a very small effect on the total flow­rate and the mean outlet temperature. This is because the resistance of the subassembly to coolant flow is already high and the additional resistance caused by a blockage is small in comparison and reduces the flow-rate only slightly. The effects of subassembly blockages are discussed in more detail later in section 5.4.1.

It may be possible, however, to detect a local blockage by observing temperature fluctuations at the subassembly outlet. A partial blockage increases the turbulence of the flow and the differences in temperature between different parts of the flow, and so causes increased temper­ature fluctuations, or “temperature noise”, at the outlet, which can be detected by fast-response thermocouples.

Multigroup diffusion calculations can normally be expected to predict the reactivity of a system within about 0.5% and the distribution of power within about 4% throughout most of the core and within 6% close to control rods or the edge of the core. The errors are due partly to uncertainties in the basic cross-section data and partly to the approximations inherent in diffusion theory. The latter explain the larger errors near interfaces between different regions.

It is possible to check calculations by means of measurements in a reactor. In the past such experimental checks were very valuable but they have become less important as the reliability of calculation methods has been confirmed.

The most useful experimental techniques are to measure reaction rates or reactivity changes due to perturbations. Fission rates can be measured with fission chambers, and for example a 238U chamber that is responsive to the flux at high energies can be combined with a 235U chamber sensitive to the flux in the keV range to assess the accuracy of a computation over the whole energy range of interest. Neutron spectra can be measured by proportional counters or time-of-flight methods. Perturbation measurements can be made by inserting small samples of various materials into the core and noting how the control rods have to be adjusted to maintain a critical balance. The control rods can be calibrated dynamically by observing the rate at which the power diverges when they are moved slightly from the critical position.

These measurements can in principle be made on power reactors but it is often inconvenient to do so partly because it is difficult to obtain access to the core, partly because the temperature and neutron flux in the core may be too high for the measuring instruments, and partly because performing reactor physics experiments conflicts with use of the reactor for the purpose for which it was built. It is very difficult to make a perturbation measurement at anything other than very low power, for example.

For this reason most, but not all, experimental checks on reactor physics calculations have come from zero power reactors. The most productive have been demountable facilities in which a reactor of almost any required composition could be assembled from samples of the various material — structure, fuel, and even sodium coolant — present in a power reactor. Test reactors of this type operated at a power of typically a few watts so that no cooling was needed. The use of zero-power experimental reactors is described in more detail by Broomfield et al. (1969).

Figure 1.6 shows calculated and measured neutron energy spectra for a zero-energy experimental assembly (Broomfield et al., 1969). The assembly was similar in composition and size to a power reactor except that it contained carbon instead of oxygen and sodium. The spectrum

was measured by a time-of-flight method, and is compared with the result of a 46-group fundamental mode calculation. The depression in the flux caused by the iron resonance at 30 keV can be seen clearly. If there had been sodium in the assembly there would have been another depression due to the important sodium resonance at 3 keV. The quantity plotted on a logarithmic scale in Figure 1.6 is the flux per unit lethargy, ф(и). “Lethargy” Uis defined by U = — logE, where E is the neutron energy. If ф(Е)dE is the flux of neutrons with energy in the range E ^ E + dE then ф(и) = 2.303Еф(Е).

The temperature difference of 1500 K or more between centre and sur­face of the fuel induces substantial thermal stresses that cause cracks. Figure 2.7 shows a cracking pattern typical of those seen in irradiated fuel. It is important to realise, however, that the pattern of cracks seen when the fuel is cool is quite different from that present during irradi­ation. The conjectured development of cracks in pellet fuel is shown
in Figure 2.10. When the fuel first generates power and experiences a temperature difference stresses are set up as indicated in Figure 2.10a. The tensile hoop stress at the outside exceeds the rupture stress and radial cracks are formed. In the centre, however, cracks are very quickly healed by the recrystallisation process.

After a certain amount of irradiation, usually about 0.1% burnup, the fuel has swollen to touch the cladding. Subsequent swelling is resisted by the cladding and the fuel is subject to a compressive stress. Under the influence of this it creeps rapidly near the centre and very slowly at the periphery. As a result the cracks close at the centre but not at the outside. When they have closed sintering or recrystallisation heals them. Thus after prolonged irradiation the centre of the fuel is free of cracks while the periphery retains cracks that tend to taper inwards (Figure 2.10b).

When the reactor is shut down the fuel cools, the centre con­tracts more than the periphery, and new radial cracks are formed that are wide at the centre and taper towards the outside (Figure 2.10c). These cracks are prominent in pictures of fuel cross-sections, such as Figure 2.7, but it should be remembered that they were not present when the reactor was operating.

EBR-1, which started operation in the United States in 1951, was a small fast reactor cooled with sodium. Its operation was largely successful and was widely reported. Possibly for these reasons it set a trend for sodium cooling, and designers of subsequent civil fast reactors kept to a coolant that was known to work rather than risk trying something else. As a result almost all fast reactors have been sodium — cooled. The only exceptions have been the USSR submarine reactors mentioned earlier for which sodium could not be used because of the impossibility of venting sodium-water reaction products.

Because there is such a preponderance of sodium design and oper­ating experience, and because at the time of writing all operating and planned fast reactors are sodium-cooled, the rest of this chapter deals only with the sodium systems of power-producing fast reactors. It con­centrates on the primary and secondary coolant circuits. The associated steam plant, apart from the steam generators, is very similar to that in a conventional fossil-fuelled power station so it is mentioned only briefly here, and the electrical equipment, which is entirely conventional, is not dealt with at all.

Chapter 3 emphasises the way in which the designer of the reactor core is constrained within very narrow limits by the properties of the materials so that there is a marked similarity between all fast reactor cores. The same is not true for the coolant circuits. The use of sodium dictates the size of the heat transfer surfaces but not the form of the circuits or the heat exchangers. There are for example two very differ­ent approaches to the layout of the primary circuit, leading to either a “loop” or a “pool” reactor. Similarly many different steam gener­ator designs are possible, employing U-tubes, straight tubes or helical tubes, with either once-through steam flow or separate evaporators and superheaters.

So far all fast power reactors have had a secondary liquid metal coolant circuit because the risk of having water and radioactive primary sodium in the same heat exchanger has been judged unaccept­able. With increasing experience of designing and operating sodium — heated steam generators, and of preventing leaks in them, the situation may change in the future, because the capital cost would be reduced if the secondary sodium circuits could be dispensed with. But because the present purpose is to describe existing and imminent systems only designs incorporating secondary sodium circuits are considered here.

It is possible to make use of the neutrons that are not needed to maintain the chain reaction in various ways. The most important is for

breeding. When a neutron is captured in 238U the 239U that is formed decays in the following way:

The times shown are the half-lives for the decay processes. As far as reactor operation is concerned the long-lived plutonium isotope 239Pu is the end-product of the chain.

239Pu has nuclear properties quite similar to those of 235U and it can be fissioned by neutrons of all energies. Neutron capture thus provides a route for converting 238U into fissile material, so 238U is called a “fertile” isotope. 232Th, which is the only naturally occurring isotope of thorium, is also fertile. It behaves very similarly to 238U: the 233Th formed on capture of a neutron decays in a chain to 233U which is long-lived and fissile.

Thus there are two naturally occurring fertile isotopes, 232Th and 238U, and three related fissile isotopes: 233U, 235U and 239Pu. There are other fissile and fertile isotopes but these five are the most important.

This ability to convert fertile isotopes to fissile raises the possibility of “breeding” new fissile material, but this can be done only if enough neutrons are available. The average number of neutrons liberated in a fission is denoted by v. Its value depends on which isotope is being fissioned and on the energy of the neutron causing the fission, but in most cases it is about 2.5. We have seen that the fact that v is greater than 1 makes a chain reaction possible: the fact that it is greater than 2 is almost equally important. If we have a reactor in which on average one neutron from each fission causes another fission to maintain the chain reaction, and if in addition more than one of the other neutrons is captured in fertile material, then the total number of fissile nuclei will increase as the reactor operates. Such a reactor is called a “breeder”.

It is sometimes said that a breeder reactor generates more fuel than it consumes. This is rather misleading. The reactor produces more fissile material than it consumes, but to do this it depends on a supply of fuel in the form of fertile material.

Although v > 2 suggests the possibility of a breeder reactor the requirement for breeding to take place is more complicated. When a neutron interacts with a fissile nucleus it does not necessarily cause fis­sion. It may be captured, and if it is, it is effectively lost. The important quantity in determining whether breeding is possible is the average number of neutrons generated per neutron absorbed. This is denoted by n, where

n = vaf Kaf + °c )■

n, which is sometimes called the “reproduction factor”, is a function of the neutron energy E, and its variation with E for the three fissile isotopes is shown in Figure 2.

Of these n neutrons one is needed to maintain the chain reaction. Some of the remainder are lost either because they diffuse out of the reactor or because they are captured by some of the other materials present, such as the coolant or the reactor structure. The others are available to be captured by fertile nuclei to create fissile nuclei. If we denote the number of neutrons lost per neutron absorbed in fissile material by L and the number captured in fertile material by C, then C is the number of fissile nuclei produced per fissile nucleus destroyed and is given by

C ~ n — 1 — L

(This is only a rough value because there are other things that may happen to neutrons).

If C is greater than one, as it must be if the reactor is to breed, it is known as the “breeding ratio”. If it is less than one it is called the “conversion ratio”. There is no logical reason for the existence of two names for C. The usage grew up because different words were used in the contexts of different reactor systems.

In practice L cannot be reduced below about 0.2, so that breeding is possible only if n is greater than about 2.2. Figure 2 shows how this can be brought about. A fast reactor using any of the three fissile materials can be made to breed, although 239Pu gives the widest margin and 235U will allow breeding only if the energy of the neutrons causing fission is not allowed to fall much below 1 MeV. In all cases the higher the neutron energy the better the breeding ratio. A 233U-fuelled thermal reactor is just able to breed but the margin is very slender. The most widely favoured breeder system is based on the use of 238U and 239Pu in fast reactors, but there is also a certain amount of interest in the 232Th — 233U system, also in fast reactors.

A fast reactor does not necessarily have to be a breeder. The excess neutrons can be used in other ways. One such is to use them to con­sume radioactive waste materials by transmutation. This process can be applied to two classes of radioactive waste: fission products and “higher actinides” (i. e. nuclides with atomic numbers greater than 94). In both cases it is the most long-lived nuclides that are of interest because of a perceived difficulty in ensuring the integrity and safety of waste storage facilities over the very long periods, up to a million years, for which the waste remains dangerous. Fission products such as 93Zr, 99Tc, 129I and 135Cs have half-lives of the order of 106 years, as does the actinide 237Np.

In most cases it is not possible to transmute any of these nuclides into stable isotopes. However, the fission products can be made less hazardous by transmuting them into other radioactive materials with shorter half-lives. Higher actinides can be eliminated by fissioning them. Some of the fissile higher actinides can be used in effect as nuclear fuel, and in all cases benefit can be taken of the energy released when they are fissioned.

It is also possible to envisage a fast reactor that, instead of breeding, acts to consume fissile material. In this way it may be possible to use a fast reactor to eliminate unwanted stocks of weapons-grade plutonium.

The effects of the temperatures of the many different parts of the reactor structure on reactivity depend on the detailed design of the reactor. The possible overall effects can be illustrated by the following examples, but the reality may be much more complex.

The radial dimensions of the reactor core are determined by the temperature of the structure that supports it while the axial dimen­sions of the core may depend mainly on the temperature of the fuel cladding, so radial and axial dimensions may change independently with temperature. If the structural temperatures increase the mean “smeared” densities of the solid materials decrease, but the coolant mean density may remain the same or it may actually increase. If, for

example, the temperatures of all the materials in the core are constant while the temperature of the structure that supports it increases so that its radius increases by 1%, the core volume will increase by 2%. The actual volume and mass of fuel and structural material in the core will remain the same, so their “smeared” densities will decrease by 2%. If the fuel and structure made up 50% of the original core, however, they will now form only 49% of the expanded core, so that the coolant volume fraction will have increased from 50 to 51%.

Figure 1.25 shows the effects of 1% increases in the radial and axial dimensions of a typical cylindrical oxide-fuelled breeder core as a function of the ratio of height to radius, H/R. In all cases the reactivity decreases with expansion because the leakage increases (in simple terms the “gaps” between the atoms are greater so the neutrons are more likely to diffuse out), but if H/ R is small the effect of increased height is small. In principle, in the limit of an infinite slab reactor (R ^ to), uniform expansion in the axial direction has no effect on reactivity at all.

1.6.3 Bowing

As the power density is higher at the centre of the core than at the edge there is a tendency for fuel elements in the outer part of the core to be hotter on the side facing the core centre. If they are straight when they are cold — i. e. when the reactor is shut down — they tend to become curved when the reactor is at power with the convex side towards the core centre. The effect on reactivity depends on how they are supported. If they were cantilevered at one end they would tend to move outwards as the power increases, reducing reactivity, but if they were held at both ends they would tend to move inwards, adding reactivity and giving positive feedback.

In practice, however, the core structure is designed so that the individual fuel elements cannot move freely at all (see Chapter 3). For this reason the effect of bowing on reactivity is likely to be small, but there may be complicated nonlinear effects as fuel elements or sub­assemblies distort to take up small clearances within the manufacturing tolerances.

The behaviour of nitride fuel is very similar to that of carbide. The thermal conductivity of UN is shown in Figure 2.24, showing that, as in carbide, high linear heat ratings are possible in principle. The conduct­ivity of PuN is about half that of UN and for mixtures containing 20 to 35% Pu values in the range of 20 Wm-1 K-1 are possible (depending on the porosity). There are fewer complications due to variation in stoichiometry because higher nitrides of uranium are unstable and can be dissociated to UN by heating to 1400 °C, and plutonium forms only PuN. The melting point of UN is about 2740 °C and PuN dissociates at about 2570 °C.

Nitride can be produced, like carbide, by carbothermic reduction of oxide, but in this case in a nitrogen atmosphere. The disadvantage of this method is that the product contains carbon and oxygen as impurities, and these pose some risk of carburisation of the cladding. Nitride is not pyrophoric.

Under irradiation in general nitride behaves similarly to carbide, but it swells less and releases less fission-product gas. It is also more prone to cracking.

The principal disadvantage of nitride fuel made with atmospheric nitrogen is the reaction 14N(n, p)14C. The 14C produced during irradi­ation is a hazard during reprocessing and is an unwanted additional component of the radioactive waste that has to be disposed of. It could in principle be reduced or eliminated by enriching the nitrogen from which the fuel is made in 15N (which constitutes 0.36% of natural nitro­gen) but this would be expensive, and it would be difficult to conserve it during reprocessing.