Nuclear Engineering and Design 239 (2009) 1406–1424

Contents lists available at ScienceDirect
Nuclear Engineering and Design
j our nal homepage: www. el sevi er . com/ l ocat e/ nucengdes
Uranium–zirconium hydride fuel properties
D. Olander
a,∗
, Ehud Greenspan
a
, Hans D. Garkisch
b
, Bojan Petrovic
c
a
Department of Nuclear Engineering, University of California at Berkeley, Berkeley, CA 94720, USA
b
Westinghouse Electric Company LLC, Pittsburgh, PA 15236, USA
c
School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA, USA
a r t i c l e i n f o
Article history:
Received 31 July 2007
Received in revised form 4 February 2008
Accepted 2 November 2008
a b s t r a c t
Properties of the two-phase hydride U
0.3
ZrH
1.6
pertinent to performance as a nuclear fuel for LWRs are
reviewed. Much of the available data come from the Space Nuclear Auxiliary Power (SNAP) program of
4 decades ago and from the more restricted data base prepared for the TRIGA research reactors some
3 decades back. Transport, mechanical, thermal and chemical properties are summarized. A principal
difference between oxide and hydride fuels is the high thermal conductivity of the latter. This feature
greatly decreases the temperature drop over the fuel during operation, thereby reducing the release of
fission gases to the fraction due only to recoil. However, very unusual early swelling due to void formation
around the uranium particles has been observed in hydride fuels. Avoidance of this source of swelling
limits the maximum fuel temperature to ∼650

C (the design limit recommended by the fuel developer
is 750

C). To satisfy this temperature limitation, the fuel-cladding gap needs to be bonded with a liquid
metal instead of helium. Because the former has a thermal conductivity ∼100 times larger than the latter,
there is no restriction on gap thickness as there is in helium-bonded fuel rods. This opens the possibility
of initial gap sizes large enough to significantly delay the onset of pellet-cladding mechanical interaction
(PCMI). The large fission-product swelling rate of hydride fuel (3× that of oxide fuel) requires an initial
radial fuel-cladding gap of ∼300m if PCMI is to be avoided. The liquid-metal bond permits operation
of the fuel at current LWR linear-heat-generation rates without exceeding any design constraint. The
behavior of hydrogen in the fuel is the source of phenomena during operation that are absent in oxide
fuels. Because of the large heat of transport (thermal diffusivity) of H in ZrH
x
, redistribution of hydrogen
in the temperature gradient in the fuel pellet changes the initial H/Zr ratio of 1.6 to ∼1.45 at the center
and ∼1.70 at the periphery. Because the density of the hydride decreases with increasing H/Zr ratio, the
result of H redistribution is to subject the interior of the pellet to a tensile stress while the outside of the
pellet is placed in compression. The resulting stress at the pellet periphery is sufficient to overcome the
tensile stress due to thermal expansion in the temperature gradient and to prevent radial cracking that
is a characteristic of oxide fuel. Several mechanisms for reduction of the H/Zr ratio during irradiation are
identified. The first is transfer of impurity oxygen in the fuel from Zr to rare-earth oxide fission products.
The second is the formation of metal hydrides by these same fission products. The third is by loss to the
plenum as H
2
.
The review of the fabrication method for the hydride fuel suggests that its production, even on a large
scale, may be significantly higher than the cost of oxide fuel fabrication.
© 2009 Elsevier B.V. All rights reserved.
1. Introduction
The history of uranium–zirconium hydride as a fuel or zir-
conium hydride as a moderator for nuclear reactors goes nearly
as far back as that of oxide fuels. This fuel/moderator occu-
pies a niche in reactor technology with a number of proposed
designs and fewer actual units. Among the former are the hydride-
moderated boiling-water superheat reactor investigated by the

Corresponding author. Fax: +1 510 526 0556.
E-mail address: fuelpr@nuc.berkeley.edu (D. Olander).
National Aeronautics Administration as early as 1960 (Gylfe et al.,
1960).
Despite the paucity and age of available data on U,Zr hydrides,
such information is essential for determining whether any
fuel-related constraints are likely to limit performance as a
power-reactor fuel. Potential limiting factors include maximum
temperature, internal rod pressure rise due to fission-gas release,
cladding strain frompellet-cladding mechanical interaction (PCMI)
and waterside corrosion.
The purpose of this contribution is to review the history of
hydride-fueled reactors in order to extract information required
for the design of a power-reactor fuel using this material and to
0029-5493/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.nucengdes.2009.04.001
D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424 1407
Fig. 1. Photomicrograph of (U
0.31
Zr)H
1.6
. The black areas are (probably) uranium
metal; the gray region is ZrH
1.6
. The white dots are unidentified (from West et al.,
1986).
summarize what is known about the properties of
uranium–zirconium hydride pertinent to its use as a fuel for LWRs.
This section reviews the work done on the early Space Nuclear
Auxiliary Power (SNAP) project starting in the 1960s, about the
same time that the popular TRIGA research reactor was marketed
to universities and nuclear laboratories. Section 2 presents most of
what is knownabout the transport, thermal, mechanical andchem-
ical properties of ZrH
x
. Section 3 gives information extracted from
30 to 40-year-old reports on the irradiation properties of hydride
fuel. A summary of these properties up to 1981 is given by Simnad
(1981). Section 4 discusses the in-reactor chemical/materials per-
formance of this material. Section 5 presents a comparison of oxide
and hydride fuels and Section 6 reviews batch production methods
of (U,Zr)H
x
. Section 7 calculates the closure of the fuel-cladding gap
as a function of burnup.
1.1. Measures of composition
As shown in Fig. 1, the hydride fuel is a two-phase mixture con-
sisting of a continuous ZrH
1.6
matrix in which small particles of
uranium metal are embedded.
Hydride fuel contains three elements, and a variety of units have
been used to denote its composition. One convenient designation
is (U
y
Zr)H
x
, where x is the H/Zr atom ratio and y is the U/Zr atom
ratio:
y =
N
U
N
Zr
(1)
where
N
Zr
=
6.02 ×10
23
¡(1 −w
U
)
91.2
(2)
¡ is the density of the fuel and w
U
is the weight fraction of uranium.
Weight fraction and atom ratio are related by
y =
91.2
M
U
×
w
U
1 −w
U
(3)
where M
U
is the atomic weight of uranium:
M
U
= 235e +238(1 −e) (4)
and e is the enrichment.The room-temperature density of the ura-
nium phase is ¡
o
U
= 19.9 g/cm
3
, and that of the zirconium hydride
phase is (Simnad, 1981):
¡
ZrHx
= (0.154 +0.0145x)
−1
g]cm
3
x - 1.6 (5a)
¡
ZrHx
= (0.171 +0.0042x)
−1
g]cm
3
x : 1.6 (5b)
Table 1
Comparison of reactor fuel features.
Reactor type SNAP LWR-hydride LWR-oxide
Weight fraction uranium in fuel, wU 0.1 0.45 0.88
Enrichment in
235
U, e 0.93 0.125 <0.05
U/Zr atom ratio in fuel, y 0.043 0.31 –
H/Zr atom ratio, x 1.68–1.83 1.6 –
Fuel density, ¡, g/cm
3
6.1 8.3 10.5
Uranium density, ¡U, gU/cm
3
0.6 3.7 9.3
235
U density=e¡U 0.56 0.46 0.46
Fuel pellet (diameter ×length, in cm) 1.4×35.6 1.2×1.5 1.2×1.5
Gap filler He, 0.1atm Liquid metal He, 20atm
Cladding material Hastelloy Zircaloy (?) Zircaloy-2
Peak fuel temperature,

C 850 550 1450
Peak linear heat rate, kW/m ∼14 37.5 37.5
Peak cladding temperature,

C ∼750 ∼350 ∼350
Average burnup 1.4×10
−3
60
FIMA
a
MWd/kgU
a
Fissions per initial metal atoms; also called “metal atom fraction fissioned”.
The density of the two-phase mixture that constitutes the hydride
is
¡ =

w
U
¡
o
U
+
1 −w
U
¡
ZrHx

−1
(6)
The uranium density of the fuel, given by
¡
U
= w
U
¡ (7)
is a key property as it dictates the enrichment of the uranium
required to achieve a desirable cycle length. The upper limit is
w
U
=0.45 (Chesworth and West, 1985).The molecular weight of the
hydride is defined as the mass in grams per mole of zirconium,
or:
M = yM
U
+91.2 +x (8)
1.2. The SNAP reactors
A hydride-fueled reactor that received more than passing atten-
tion was developed for the SNAP program at Atomics International
under the auspices of the Atomic Energy Commission, the prede-
cessor of the current Dept. of Energy (Lillie et al., 1973). Six reactors
with thermal outputs ranging from 50kW to 1MW were built and
operated, and one was placed in earth orbit. A substantial body of
experimental irradiation tests was reported in this program, many
of which have a bearing on the hydride-fueled LWR that is the sub-
ject of this issue. Table 1 compares the pertinent characteristics
of three reactor types: the SNAP reactor, the LWR-hydride reactor
(Shuffler et al., this issue-a) and the LWR-oxide reactor (a standard
BWR).
Noteworthy in Table 1 is the wide range of uranium contents of
the fuels. Hydride fuels operate most reliably withlowUconcentra-
tions, which is the reason for the choice of 10wt% U, corresponding
to a uranium density of 0.6gU/cm
3
, in the SNAP reactors. To pro-
duce acceptable amounts of nuclear energy with so little uranium
per unit volume, the enrichment is 93%
235
U. In order to attain
the same
235
U density of the fresh fuel as in an oxide-fueled core,
the enrichment of the LWR-hydride fuel would have to be scaled
according to the total uranium density, or:
e(LWR −hydride) = e(LWR −oxide) ×(9.3]3.7).
Tomatcha 5%-enrichedoxide fuel inthis regard, the hydride fuel
for LWR use would require 12.5% enriched uranium. This enrich-
ment is higher than the current regulatory limit of 5%, but is below
the LEU limit of 20%.
The power density of SNAP fuel was only ∼1/4 that of current
LWR oxide fuel. But then, the SNAP reactors were not designed
1408 D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424
for power production. They did, however, operate at very much
higher fuel and cladding temperatures than the proposed hydride
LWR fuel. At 850

C fuel centerline temperature and the H/Zr values
listed in the table for the SNAP reactor, the equilibrium H
2
pres-
sure is ∼3atm. While this pressure does not threaten the cladding
mechanically, it risks hydrogen loss by permeation through the
metal. This phenomenon was the major performance constraint for
these reactors, and is in part the reason for the choice of Hastelloy
cladding for SNAP reactor fuel.
LWRs cannot tolerate Hastelloy as cladding because of its high
nickel content andconsequent negative effect onneutroneconomy.
Also, LWR hydride fuel cannot operate for a long period at >800

C
becauseof theassuredcatastrophic reactionof theZircaloycladding
with gaseous hydrogen inside the fuel element. Fortunately, the
hydride fuel peak temperature at the LWR power density is only
about 550

C (Shuffler et al., this issue-b).
Given the high
235
U content of SNAP reactor fuel, high bur-
nup was easily attainable. The maximum burnup of the oxide fuel
(60MWd/kgU) is 1/5 that of the SNAP hydride fuel.
1.3. TRIGA reactors
The most widely used of hydride-fueled reactors is the well-
known TRIGA developed by General Atomics (Simnad, 1981;
Chesworth and West, 1985). With a steady-state power of a few
MW, this reactor is primarily for researchpurposes. Early TRIGAfuel
utilized 8wt% of highly enriched uranium. With the restriction on
enrichment to 20%, the proportion of uraniumin the fuel increased
to 45wt%, or 21 volume percent. Even at this loading, the uranium
density in the hydride fuel is only 40% of that in UO
2
. Increasing the
235
U content is the only way of overcoming this deficiency. TRIGA
cores are cooled by natural convection with water at about 60

C.
Compared to current oxide power-reactor fuel, TRIGA fuel is very
large: a single “pellet” is about 4cm diameter and 35cm long. The
cladding is either aluminumor stainless steel. All proposed or con-
structed hydride-fueled reactors have adopted the basic TRIGAfuel.
The exception to the usual low power of these research reactors
is the Romanian TRIGA, which utilizes standard-size LWR fuel pel-
lets made of (U,Zr)H
1.6
instead of UO
2
. This hydride fuel is reported
to have operated at a linear heat rate (LHR) of ∼80kW/m at a fuel
centerline temperature of 820

C with a forced-convection coolant
temperature of 60

C (Iorgulis et al., 1998; Toma et al., 2002). The
design limits set for the high-power TRIGA core are (Iorgulis et al.,
1998; Toma et al., 2002) fuel temperatures of 750

C at steady-state
and 1050

C during transients. The proposed hydride-fueled LWR is
less thermally- demanding than the high-power TRIGA.
2. (U,Zr) hydride properties
2.1. Hydrogen diffusion
Unlike oxide fuel, where diffusion of corrosion-product hydro-
gen in ␣-Zry is a key transport property, the other two major
phases (-Zry metal and in the ceramic -ZrH
x
) are important for
hydride fuels. Hydrogen diffusion in the metal controls the kinetics
of hydriding during fabrication and in the hydride phase, hydro-
Fig. 2. Diffusivity of H in ␤-Zr.
Fig. 3. Diffusivity of H in ␦-hydride.
gen mobility dictates the kinetics of hydrogen redistribution in the
temperature gradient during operation.
The hydrogen diffusivities in these two phases are shown in
Figs. 2 and 3. While the agreement among the measured diffusivi-
ties in ␤-Zr is fair (maximumof a factor of 10 discrepancy), the data
in Fig. 3 exhibit differences of a factor of 50 between the lowest and
highest lines. Table 2 provides additional details on hydrogen diffu-
sion in the ␦-hydride. The two early measurements were obtained
using the classical method of absorption of hydrogen from H
2
into
a disk or cylinder specimen of ZrH
x
. In this method, diffusivities are
determined from curves of weight gain versus time or progression
of the delta phase into the metal. The latest measurement by Majer
et al. (1994) utilized nuclear magnetic resonance, which detects
the mean residence time (z) of a proton on an interstitial site in
the lattice. This measurement yields the activation energy for dif-
fusion (E
d
, last column in Table 2). Related techniques yield the
mean jump distance (L), which corresponds to the separation of
tetrahedral interstitial sites in the fcc structure of the Zr atoms in
the crystal. The diagram of the structure of the ␦-hydride crystal in
Table 2
Diffusion coefficients in zirconium hydride.
Authors Year Method
a
H/Zr range T range,

C Do, cm
2
/s ED/R, K
Albrecht and Goode (1960) 1960 Diff 1.56–1.86 500–750 6.0×10
2
17600
Paetz and Lücke (1971) 1971 Diff 1.50–1.70 650–800 2.5×10
−1
8960
Majer et al. (1994) 1994 NMR 1.58–1.86 330–700 1.5×10
−3b
7100
b
a
Diff: rate of absorption of hydrogen by solid; NMR: nuclear magnetic resonance.
b
For H/Zr =1.58.
D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424 1409
Fig. 4. Crystal structure of zirconium hydride. The zirconium atoms (large, gray)
form a face-centered-cubic crystal lattice. The H atoms (small sphere) occupy tetra-
hedral sites, one of which is outlined in the sketch.
Fig. 4 shows that there are 8 tetrahedral interstitial sites in the unit
cell, which consists of 4 zirconiumatoms (8 corner atoms shared by
8 contiguous unit cells and 6 face-centered atoms shared between
two adjoining unit cells).
The hydrogen diffusion coefficient the delta hydride is given by
D
H
=
jL
2
6z
The factor f is a correlation factor; it accounts for the probability
that the jumping atomreturns to the site whence it came. It is mod-
eratelyless thanunity. Themeanresidencetimez depends uponthe
jump frequency ¡[exp(−E
D
/RT)], where ¡ is the vibration frequency
of the H atom in the interstitial site and E
D
is the potential-energy
barrier height separating sites, the number of nearest-neighbor
interstitial sites (6) and the probability that a site is occupied (x/2,
where x represents the H/Zr ratio).
1
z
= 6

1 −
x
2

¡

exp


£
D
RT

The NMR method measures the diffusivity at a single concen-
tration. Based on this feature, Majer et al. (1994) showed that the
measured concentration dependence of the pre-exponential factor
closely follows the theoretical expectation:
D
H
=

jL
2

1 −
x
2

¡

exp


£
D
RT

(9)
The quantity in the square brackets is the pre-exponential factor
D
o
. In the range of the experiments by Majer et al. (1994) shown
in Table 2, the activation energy for diffusion (divided by the gas
constant) is a weak function of H/Zr ratio, increasing from 7100K
at H/Zr =1.58 to 7800K at H/Zr =1.86. This feature may be due to
incursions of the ␧-hydride in the sample.
2.2. Thermal diffusivity
The other transport property of ZrH
x
that affects fuel perfor-
mance is the thermal diffusivity(or heat of transport). This property
causes hydrogen in the Zr matrix to move towards cold regions,
evenintheabsenceof aconcentrationgradient. Theprocess reaches
steady state when the gradient due to concentration diffusion just
Fig. 5. Hydrogen redistribution in ZrH
1.6
with helium or liquid metal in the fuel-
cladding gap; To and T
S
are the centerline and surface temperatures, respectively
(Olander and Ng, 2005).
balances that due to thermal diffusion. In cylindrical coordinates,
this condition of zero net flux is given by
dx
dr
+
1
Q
1
2
x
d1
dr
= 0 (10)
where T is the temperature (in K), x is the H/Zr ratio, r is the radial
position in the fuel pellet and T
Q
is the heat of transport of H in
ZrH
x
divided by the gas constant. The only known measure of this
quantity is that due to Sommer and Dennison (1960), who report
T
Q
=640K.
The solution to Eq. (10) is (Huang et al., 2000; Olander and Ng,
2005):
x = /e
1
Q
/1
(11)
The constant of integration A is determined from the specified
average H/Zr ratio of the fuel:
x
avg
=
2
R
2

R
0
rx(r) dr (12)
Fig. 5 shows the extent of hydrogen redistribution in ZrH
1.6
in a
parabolic temperature distribution. The solid curve is for a helium-
bondedfuel element andthe dashedcurve is for a fuel element with
a liquid metal in the fuel-cladding gap. Redistribution is extensive;
the original H/Zr ratio of 1.6 has diminished to 1.46–1.49 at the
center of the pellet and grown to 1.74–1.78 at the periphery. This
redistribution has only a minor effect on the neutronic characteris-
tics of hydride fuel.
2.3. Thermodynamic properties
The most important thermodynamic properties of the Zr–Hsys-
tem are the phase diagram and the equilibrium H
2
pressure as a
function of H/Zr and temperature. These two properties are con-
tained in Fig. 6, which shows isobars of the equilibrium hydrogen
pressure superimposed on the binary phase diagram (Zuzek and
Abriata, 1990; Wang and Olander, 1995). At temperature exceeding
∼800

C, theequilibriumH
2
pressureover thehydrideis sufficiently
large that accumulation of this gas in the plenum or hydriding of
the cladding would occur, accompanied by a reduction of the H/Zr
ratio of the fuel. If all released H2 migrated to the plenum, the
neutron-moderating capability of the fuel would be reduced.
2.4. Thermal and mechanical
Table 3 summarizes the principal thermal and mechanical prop-
erties of the delta hydride.
1410 D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424
Fig. 6. H-Zr phase diagram with isobars of equilibrium H
2
pressures superimposed (Huang et al., 2000; Olander and Ng, 2005).
Since the uranium metal in the hydride has a conductivity of
∼28W/m-K, the two-phase thermal conductivity is ∼20W/m-K.
This is about 7 times that of UO
2
.
To reduce the fuel centerline temperature in hydride fuel, a low-
melting, inert liquid metal (Pb–Bi–Sn) replaces helium as the gap
filler. This replacement lowers the centerline temperature by well
over 100

C (Wongsawaeng and Olander, 2007).
The combination of the density variation with H/Zr ratio (2nd
rowof Table3), thermal expansion(1st rowof Table3) andhydrogen
redistributionduetothein-reactor temperaturedistribution(Fig. 5)
results in peculiar mechanical behavior of hydride fuel on startup
(Olander and Ng, 2005). In the parabolic temperature distribution
engendered by fission, thermal expansion results in azimuthal ten-
sile stresses at the pellet periphery that exceed the fracture stress
(barelyor quiteabit, dependingonwhichvalueof thefracturestress
in Table 3 is accurate). The combination of hydrogen redistribution
towards thecool peripheryof thepellet andexpansionwithincreas-
ing H/Zr ratio results in significant compressive stresses, called
“hydrogen stresses” in the outer portion of the pellet. This phe-
nomenon was first noted (but not analyzed) by Gylfe et al. (1960),
where it was termed “migration stress”.
The total stress is the sumof the thermal and hydrogen stresses.
These two axial components and their sum are shown in Fig. 7 for
typical liquid-metal-bonded fuel-pellet operating conditions. Axial
stresses are shown because this component is larger than the total
azimuthal stress. According to Table 3, the fracture stress lies some-
where between the dotted lines in the figure. The thermal stress
actingaloneis capableof crackingthepellet at thesurface. However,
the hydrogen stress is highly compressive at large radii, and more
Fig. 7. Axial stresses in a LM-bonded hydride fuel rod (Olander and Ng, 2005).
than overcomes the tensile thermal stress. The total axial stress is
tensile out to a fractional radius of 0.8 and exceeds the range of the
fracture stresses over most of this interval.
However, the thermal andhydrogenstress components have dif-
ferent time responses. The thermal stress develops as soon as the
steady-state temperature distribution in the pellet is established.
This occurs in a couple of days over which reactor startup takes
place. On the other hand, the kinetics of hydrogen redistribution,
which is controlled by the diffusivity of hydrogen in the hydride
Table 3
Thermal and mechanical properties of (U
0.31
,Zr)H
1.6
.
Property Value Reference
Thermal expansion coeff.,

C
−1
˛=7.4×10
−6
(1+2×10
−3
T)
a
Simnad (1981), Beck and Mueller (1968), Yamanaka et al. (1999)
Hydrogen expansion (^L/L)
H/Zr
=0.027(H/Zr–1.6) Gylfe et al. (1960), Simnad (1981)
Young’s modulus, GPa 130 Simnad (1981), Yamanaka et al. (1999)
Poisson’s ratio 0.32 Simnad (1981)
Fracture stress, MPa 200 (ten.), 55 (ten.), 100 (comp.), Merten et al. (1958), Gylfe et al. (1960), Beck and Mueller (1968)
Thermal Conductivity, W/m-K 18±1 Yamanaka et al. (2001)
Heat capacity, J/mole-K (25+4.7x) +(0.31+2.01x)T/100+(1.9+6.4x)/T
2
×10
−5a
Yamanaka et al. (2001)
a
300<T (K) <1000; x =H/Zr.
D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424 1411
solid are slower. The transient hydrogen redistribution calculations
by Huang et al. (2000) show a strong dependence upon the lin-
ear heat rate: at 30kW/m, steady state is achieved in ∼75h, but at
15kW/m, ∼10 days are required. Interpretation of curves such as
those shown in Fig. 7 must take into account the possible differ-
ences in the time constants of the thermal- and hydrogen-stress
components.
2.5. Chemical properties
Chemical interaction between zirconium hydride and Zircaloy
cladding is thermodynamically favored but the kinetics are
unknown. Hence, designation of Zircaloy as the cladding metal for
hydride fuel rods must be approached with great caution. First,
consider the worst case.
The surface temperature of a helium-bonded hydride fuel pel-
let operating at a linear heat rate of 37kW/m with a 35-␮m gap
thickness is ∼500

C (Table 5). Taking into account the redistribu-
tion shown in Fig. 5 and the phase diagramof Fig. 6, the fuel surface
generates an equilibrium H
2
pressure of 5×10
−3
atm. This partial
pressure is not sufficiently large to move hydrogen along the gap to
the plenum, but it can drive hydrogen to the cladding inside surface
where metallic zirconium essentially reduces p(H
2
) to zero. Means
of avoiding this possibility are discussed below.
Replacing heliumin the gap with a liquid metal prevents hydro-
gen from reaching the cladding inner surface as long as the gap is
open. However, closure of the fuel-cladding gap generates pellet-
cladding mechanical interaction (PCMI). Hard contact of the two
solids may provide a mechanism to overcome the sluggishness of
the gas-surface reaction. The conservative design response to this
situation is to avoid PCMI entirely by making the initial gap large
enough that the combination of fuel swelling and cladding creep-
downdoes not closethegapinthelifetimeof thefuel. Aninitial cold,
radial gap of ∼150␮m(compared to ∼80␮min conventional oxide
fuel elements) is sufficient to prevent PCMI in BWR hydride fuel to
a burnup of 60MWd/kgU. With LM-bonding, no fuel-temperature
increase results from the larger gap size.
Other methods have been suggested to separate the fuel and
Zircaloy cladding. Most of them involve adding a hydrogen perme-
ation barrier to the cladding’s inner surface. They include:
2.5.1. Stainless steel liner
The transition metals do not form stable hydrides, and the sol-
ubility of hydrogen in stainless steel (SS) is low (but not zero).
To exploit this property, duplex cladding consisting of a Zircaloy
outer tube and a SS inner liner has been considered as a means of
reducing hydrogen loss fromthe fuel. Analysis of the hydrogen per-
meation process is given in Appendix A. Using the permeability of
hydrogen in stainless steel reported by LeClaire (1984), 17 years are
required to reduce the H/Zr ratio of the fuel from 1.6 to 1.55. From
a hydrogen-retention point of view, the SS liner is very effective.
The advantages of the liner over cladding entirely made of stain-
less steel are
(a) Reduced parasitic neutron capture by the elements in the alloy.
(b) Avoidance of waterside intergranular stress-corrosion cracking
(IGSCC) of the SS.
The primary disadvantages of this approach include:
(a) In BWR cladding, the absence of the soft Zr liner for alleviating
cracking by pellet-cladding interaction (PCI).
(b) The SS liner may also be susceptible to internal stress corrosion
cracking by fission products.
(c) Differential expansion between Zr and SS may cause debonding
of the two.
2.5.2. SiC internal coating or sleeve
Silicon carbide is very resistant to chemical attack by or perme-
ation of hydrogen; neutronically, it is practically inert and is stable
against radiation damage to very high neutron fluences. It is used
very successfully in HTGR fuel as a hydrogen (tritium) permeation
barrier (Greenspan, 1998).
2.5.3. Glass-enamel coating
A glass-enamel coating metal cladding, about 0.08mm thick,
has been successfully bonded to the inner surface of the Hastelloy
cladding in SNAP reactors and survived operation at temperatures
up to 700

C (Hesketh and Stanbridge, 1993).
2.5.4. Zirconia coatings
The solubility of H
2
in ZrO
2
is very low (several ppm atomic);
consequently the oxide should be very resistant to hydrogen per-
meation. From the point of view of fabrication, oxidizing the inner
surface of Zircaloy cladding is the simplest, and certainly the
least expensive, of all coating methods. However, its effectiveness
depends on its resistance to cracking, which afflicts the waterside
corrosion scale on the cladding outer surface.
Oxidizing the outer surface of the hydride fuel pellets has been
demonstrated (and patented, Eggers, 1978) and appears to provide
a satisfactory hydrogen permeation barrier up to 800

C. In addi-
tion, when the fuel-cladding gap closes, the oxide layer between
the hydride and Zircaloy prevents direct contact between the latter
two solids.
3. Irradiation effects
In common with all nuclear fuels, the key irradiation properties
are fuel swelling, fission gas release and fission-product chemistry.
These are functions of temperature (or linear heat rate) andburnup.
3.1. Burnup units
The literature on hydride fuels expresses burnup either as F, the
number of fissions per unit volume, or FIMA, “Fissions per Initial
Metal Atom”
1
:
FIMA =
F
(N
U
+N
Zr
)
(13)
N
U
and N
Zr
are the atom densities of U and Zr in the as-fabricated
fuel, respectively. N
Zr
is given by Eq. (2).
In power reactor literature, on the other hand, burnup is
expressed as MWd/kgU, where the mass of U refers to fresh fuel.
However, this unit is not appropriate for comparing hydride and
oxide fuels, as can be seen from the following. With the energy of
fission taken as 3.2×10
−11
J/fission, the MWd/kgUunit of burnup is
related to the number of fissions per cm
3
and the uranium density
of the fresh fuel by
BU

MWd
kgU

= 3.7 ×10
−19
F
¡
U
(14)
Conversion from FIMA to BU is accomplished with:
BU =
3.7 ×10
−19
¡
U
×N
Zr
(1 +y) ×FIMA
= 2.4 ×10
3

1 −w
U
w
U

(1 +y) ×FIMA (15)
1
%FIMA is equivalent to the burnup unit “metal atom %” that is commonly used
in the literature on hydride fuel.
1412 D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424
As shown in Table 1, the uranium densities of the hydride fuels
are substantially lower than that of oxide fuel. If oxide and hydride
fuels are to be compared on the basis of energy output per unit fuel
volume (which is proportional to F), the uranium density should
be removed from Eq. (14). A more appropriate definition for this
purpose is the “equivalent oxide” burnup:
BU
equiv.ox.
= BU×
¡
U
9.3
= 4.0 ×10
−20
F (16)
where 9.3 is the density of uranium in UO
2
. Fuels with the same
equivalent-oxide burnup have generated the same amount of
energy per unit volume.
Because FIMAis commonly used for hydride fuels, conversion to
the equivalent-oxide burnup is required. This is accomplished with
the aid of Eq. (15):
BU
equiv.ox.
= 2.4 ×10
3

¡
U
9.3

1 −w
U
w
U

(1 +y) ×FIMA (16)
For the 45Uwt% hydride considered here, Eq. (16) reduces to:
BU
equiv. oxide
= 1.5 ×10
3
×FIMA (16a)
Thus, 1% FIMA in this particular hydride fuel is equivalent
(in terms of energy production per unit volume) to a burnup of
15MWd/kgU in UO
2
.
3.2. Fuel Swelling due to void formation
Very early in life, oxide fuel shrinks (densifies) because of
destruction of fabrication porosity by fission fragments. The
early-time volume change of hydride fuel is either nonexis-
tent (T <∼700

C) or exhibits large increases (T >700

C) termed
“offset swelling”(Lillie et al., 1973). Following these initial
swelling/densification stages, both fuels continue to swell linearly
with burnup as a result of introduction of fission products into the
matrix.
In the SNAP program(Lillie et al., 1973), swelling of hydride fuel
was measured both in-reactor and by quite sophisticated irradia-
tion tests. In normal irradiation tests, the effects of temperature
and fission rate cannot be separated because the latter controls the
former. The SNAP program utilized a test rig in which cooling of
the in-reactor specimens was accomplished by a gas flowing along
the exterior of the specimen holder. Without changing the fission
rate, the specimen temperature was controlled by using different
mixtures of helium and xenon in the coolant gas, thereby enabling
examination of the effect of each variable separately. Swelling
was determined by density measurements of the irradiated
specimens.
Microscopic examination of swollen hydride fuel revealed large
cavities adjacent to uranium particles. These cavities are believed
to be true voids, not fission-gas bubbles. Voids are generated in the
hydride matrix close to the uranium particle by the radiation dam-
age created by escaping fission fragments. Fig. 8 is a transmission
electron micrograph of these voids (Lillie et al., 1973). The ura-
nium particles are not imaged in this TEM, but were reported as
∼5␮m in diameter and ∼40␮m apart. The temperature of the fuel
at the location of the TEM was estimated at 785

C and the burnup
was ∼2×10
−3
FIMA, which is about twice the saturation burnup
for offset swelling. The void-size distribution was roughly bell-
shaped, ranging from the microscope’s resolution limit of 10nm
up to a maximum diameter of ∼120nm. The average diameter (not
defined) was about 60nm. The number density of these voids, aver-
aged over an area comprising numerous uranium particles, was
∼5×10
−7
nm
−3
. From these figures, the overall volume swelling
should be:

^v
v

void
=
¬
¯
D
3
6
N
void
=
¬ ×60
3
6
×5 ×10
−7
= 0.06
Fig. 8. Voids near a uranium particle following irradiation of hydride fuel in a SNAP
reactor (Lillie et al., 1973).
The (^V/V) values reported by Lillie et al. (1973) were ∼10% in
the immediate vicinity of the uranium particles and ∼4% averaged
over a large cross-section of the fuel. The swelling calculated by the
above equation lies between these two observed values.
Fig. 9 shows that offset swelling can be quite large and highly
temperature-sensitive: at 760

C, it is 5%, but it vanishes at ∼650

C.
In these plots, the offset swelling saturates at a burnup of 10
−3
FIMA; however, no experimental evidence was offered to support
this supposition.
The top graph in Fig. 9 shows unusual changes in offset swelling
with the fission rate, or the linear heat rate. The decrease in offset
swelling with this variable is attributed to behavior of voids as the
primary source of the volume change.
The data underlyingthe plots inFig. 9are showninFig. 10. Unfor-
tunately, the burnup rates for each data point are not given by Lillie
et al. (1973). The lines shown in the graph were estimated from
offset swelling values taken from Fig. 9. They fit the equations:
log

%^v
v

void
= / −8

10
3
1

(17)
The coefficients A and B depend on the burnup rate according
to:
/ = 9.2 +2.2 ×10
3
(FIMA]year)
8 = 8.7 +2.4 ×10
3
(FIMA]year) (18)
Comparison of the void production in irradiated hydride reac-
tor fuel with the well-developed theories of void swelling in metals
provides some insight into the behavior shown in Fig. 9. In par-
ticular, the features for comparison are the dependence of void
size/swelling on temperature and burnup rate and the conditions
that lead to saturation of void growth.
3.2.1. Temperature dependence
Fig. 9 shows a sharply increasing offset swelling with increasing
temperature. The typical bell-shaped curves of void-growth rate
D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424 1413
Fig. 9. Swelling behavior of (U,Zr)Hx SNAP fuel.
vs temperature in metals (e.g., Fig. 19.15 of Olander, 1976) exhibit
similar behavior at the onset of swelling. However, the temperature
effect observedinhydridefuel is onthemagnitude of offset swelling,
while the temperature effect seeninmetals refers to the rate of void
swelling.
3.2.2. Burnup rate
The dramatic effect of the burnup rate on offset swelling in
hydride fuel (Fig. 9) does not have a direct counterpart in void
Fig. 10. SNAP fuel-offset void-swelling data (from Lillie et al., 1973).
swelling of metals. The simple theory of the latter (e.g., Section
19.5.11 of Olander, 1976) predicts a persistent, slow void-radius
growth rate of the form
˙
R ∝

dpa/s, where dpa/s is the rate of
atomic displacements. If this behavior also implies an increase in
thesaturationvoidsizeinmetals withdamagerate, theburnup-rate
effect in hydride fuel is diametrically opposite.
3.2.3. Saturation of swelling
The offset swelling values plotted as data or correlation lines
in Fig. 10 are (supposedly) independent of burnup. However, it
is doubtful that the experiments reported by Lillie et al. (1973)
accurately measured the burnup at which saturation occurred.
It is unlikely that saturation occurred at 10
−3
FIMA for all tem-
peratures and burnup rates, as shown in Fig. 9. Nonetheless, it
appears clear that saturation of offset swelling occurred, so a
comparison with the analogous phenomenon in metals is appro-
priate.
Two theories explaining saturation of void growth in metals are
summarized in Section 19.5.15 of Olander (1976), which contains
citations of the original sources. However, there is no experimental
evidence for saturation of void swelling in metals.
The first model visualizes a stationary state with all voids con-
nected by curved dislocations, with no free dislocations left to
climb. This situation just balances the preference of interstitials for
dislocation lines, which is the root cause of void swelling in metals.
Because all dislocation lines are pinned by the voids, they can no
longer climb by absorbing interstitials, even though the bias factor
difference favoring interstitials remains. The quantitative develop-
ment of this model yields a functiongiving the combinations of void
radius, void number density, and dislocation density at which void
growth ceases.
The second model examines the effect of dislocation loops near
voids. The dislocation creates a stress field around itself. In order
to maintain a stress-free void surface, an induced, or image, stress
field is created in the void. The void in essence now behaves like a
dislocation, and acquires the same bias for interstitials as the actual
dislocations. Because both sinks for point defects are equally biased
for interstitials, all point defects generated by irradiation are either
annihilated by recombination or flow in equal numbers to voids
and dislocations. As a result, void growth ceases.
It is conceivable that saturation of void growth in hydride fuel
occurs because of one of the above metal-based models. However,
all models for metals are founded on the creation of damage by
fast neutrons whereas the principal atom-displacement agent in
irradiated hydride fuel is the fission fragment. Fission fragments
can destroy voids or intragranular bubbles in UO
2
by a mecha-
nism termed “re-solution”. A brief review of the literature on this
subject is given by Olander and Wongsawaeng (2006). A com-
mon observation in irradiated UO
2
is a population of intragranular
bubbles whose size is independent of burnup. This observation is
explained by a balance between bubble growth by absorption of
gas atoms (and vacancies) generated by fission in the solid lattice
and bubble destruction by fission-fragment re-solution. Although
this model is advanced for cavities containing fission gas, there
is no reason that it could not be applied to voids in hydride
fuel.
2
3.3. Fuel swelling due to fission products
For UO
2
, fuel dimensional changes during irradiation are due
principally to the ingrowth of solid fission products, with smaller
2
Lillie et al. (1973) assume that the cavities observed in irradiated hydride fuel
were true voids, but the possibility that the cavities contained fission gas was not
investigated.
1414 D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424
contributions due to fission-gas bubbles and early-on, to removal
of as-fabricated porosity (densification). The solid fission-product
swelling rate is about 0.07% per MWd/kgU.
In hydride fuel, the fission-product-swelling rate is represented
by the slope of the lines in Fig. 9 after cessation of offset swelling.
Following this period, hydride fuel swells at a constant rate of ∼3%
per %FIMA. Converting fromFIMAto equivalent-oxide burnup gives
a swelling rate of 3/1.5×10
3
=0.002, or
^v
v
= 0.2% per BU
equiv. ox
On this basis, the fission-product swelling of hydride fuel is ∼3
times larger than that of oxide fuel. It is also twice as large as the
value computed for the contribution of solid fission products by
Huang et al. (2001). A good part of the reason for this significant
difference is the presence of Xe +Kr, Te and I in the fuel, which were
not taken into account by Huang et al. (2001). Below∼750

C these
volatile elements are retained in the fuel (see Section 3.4). The rare
gases may exist as nanobubbles that were not observed in the SNAP
studies.
Fission-product swelling has a significant effect on closure of
the fuel-cladding gap and the concomitant onset of pellet-cladding
mechanical interaction (PCMI) (see Section 7).
3.4. Fission-gas release
A major difference between oxide and hydride fuels involves
the mechanisms responsible for the release of fission gas during
irradiation. This phenomenon has been exhaustively studied for
uranium dioxide fuel, and there is enough empirical evidence to
support the predictions of analytical and computational models.
However, the data on fission-gas release from hydride fuel lack
robustness and careful validation. What data are available sug-
gest that recoil is the only mechanism of fission-gas release up
to temperatures in the neighborhood of 700

C (Chesworth and
West, 1985). The release fraction due to recoil is of the order of
10
−4
, supporting the assumption that fission gas is not released
from hydride fuel. The justification for this statement is presented
below.
Many ways are available for quantitative measurement of
fission-gas release from irradiated fuel.
In-reactor tests include measurement of:
1. the buildup of pressure within an instrumented test fuel rod
resulting from the release of the stable Xe isotopes (
132
Xe,
134
Xe
and
136
Xe) and
85
Kr. The released gas accumulates in the plenum
and the fuel-cladding gap. The fraction released increases with
time.
2. the rate of release of a short-half-life isotope (
85m
Kr or
131
Xe) by
passing a carrier gas through a test fuel rod or fuel specimen. The
released radioisotopes are trapped on refrigerated charcoal and
then measured by gamma spectroscopy. Steady-state release is
achieved and the ratio of the release rate (R) to the birth rate (B)
(i.e., the production rate in the fuel) is determined. This R/B ratio
is often (erroneously) called the fraction released.
The post-irradiation test involves:
3. annealing the specimen in a furnace through which a carrier gas
flows. Typically,
133
Xe gamma activity is used to measure the
quantity released. For such an experiment the fraction released
is the ratio of the amount of nuclide released in a certain time
divided by the quantity of the isotope contained in the specimen
following prior irradiation.
Interpretation of these experiments ranges from the very sim-
plest Boothmodel to partially mechanistic models included inlarge
fuel-behavior codes. The Boothmodel is basedondiffusionfromthe
interior of the grains of the fuel to the grain boundaries, whence
release occurs (see Section 15.4 of Olander, 1976). The basic proper-
ties that govern the interpretation of all three of the tests described
above are the diffusivity of the isotope in the grains (D) and the
radius of the grains (a), which appear as the combination a
2
/D. This
parameter is the onlyone neededinthe Boothmodel andappears in
different forms inthe quantitative interpretationof the tests. What-
ever model is used to analyze test data, the expectation is that the
same model and its parameters can be used to accurately predict
releaseunder power reactor conditions, anexpectationthat is rarely
met.
All fission-gas release theories, from the Booth model on, were
developedfor applicationtoUO
2
fuel. Becausethemicrostructureof
U–Zr hydride in no way resembles that of UO
2
, with one exception,
the models are useless for understanding release from the former
fuel. The exception is the mechanism by which fission fragments
born within ∼10␮m of an exposed surface escape from the solid.
This so-called recoil mechanism produces a release rate-to-birth
rate ratio equal to 1/4q(S/V)¡, where S/V is the surface-to-volume
ratio of the fuel, ¡ is the range of fission fragments in the fuel and
q is the fraction of the escaping fission fragments that are not re-
implanted in the opposite surface. Typical values of recoil R/B are
10
−4
.
There are numerous reports of fission product release from
U–Zr hydride (TRIGA) fuel (Chesworth and West, 1985; West
et al., 1986; GA Technologies, unknown). Unfortunately none of
them are the original sources of the data they report. A com-
pendium of these second-hand reports is shown in Fig. 11. This
plot appears in all reports of fission-gas release from hydride
fuel, with the latest experimental results added to it. Most of the
results come from post-irradiation anneal experiments, with a few
experiments of the in-reactor type 2 described above. No data
have been obtained from in-reactor fuel rods (method 1 above),
the claim being that no fission gas isotopes could be detected in
post-irradiation examination. How seriously to take this claim is
uncertain, since the temperature of the fuel was poorly known or
unknown.
The dashed horizontal line in Fig. 11 is attributed (by the cited
references) torecoil release, which, giventhe magnitude of the frac-
tion release and its temperature independence up to 600–800

C, is
a reasonable interpretation. The curve through the remaining data
points is not model-based. The only claim is that it is due to some
undefined “diffusion” process.
Two problems are immediately evident from Fig. 11.
First, release fractions measuredby post-irradiationanneal tests
are reportedas a functionof temperature only. However, the release
Fig. 11. Fission-gas release from U–Zr hydride fuel (Olander and Ng, 2005).
D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424 1415
Table 4
Probable chemical and physical states of fission products in oxide and hydride fuels.
Element
a
Yield States in oxide States in hydride
Chemical Physical Chemical Physical
Zr (Nb) 0.3 ZrO
2
Solution
b
ZrH
1.6-x
␦-hydride
RE
c
(Y) 0.53 RE
2
O
3
Solution REH
2
In ␦-hydride?
Ba (Sr) 0.15 BaZrO
3
Oxide #1 (Ba,Sr)H
2
In ␦-hydride?
Cs (Rb) 0.2 Cs, Cs
2
UO
4
, CsI Gas, oxide #2 Cs, CsI Gas, solid
Mo 0.25 MoO
2
, Mo Solution, alloy ? ?
NM
d
0.25 Elements Alloy U(NM)
3
Intermetallic
I (Te) 0.01 CsI Compound CsI Solid
Xe (Kr) 0.25 Element Gas Element Gas
U (Pu) – UO
2
Oxide U Element
a
The element in parentheses is closely related chemically and physically to the first element.
b
Dissolved in UO
2
.
c
RE=rare earths (La, Ce, Pr, Nd, Sm, Pm, Eu, Gd).
d
NM=noble metals (Pd, Ru, Rh, Tc, Pt).
fraction from this type of test is a function of time as well as
temperature.
3
Second, thelow-temperaturedatapoints inFig. 11werereported
to have been determined by the post-irradiation technique. How-
ever, recoil release can only be detected while fission is occu-
rring.
Despite the rudimentary state of understanding of the mecha-
nism(s) of fission-gas release from hydride fuel, the available data
strongly suggest that if fuel temperatures are kept below ∼650

C
during operation, fission-gas release is negligible. In the article by
Shuffler et al. (this issue-b), the design limit for the centerline tem-
perature at steady-state is 750

C. In view of the observed swelling
behavior, this value may be somewhat aggressive. It is essential that
irradiation tests of hydride fuel be conducted in order to verify the
behavior shown in Fig. 11.
3.5. State of fission products
The chemical andphysical states of the fissionproducts inirradi-
atedU/Zr hydride fuel has beenstudiedby Huang et al. (2001) using
the thermochemical code Thermo-Calc. Mixtures of the elements U,
Zr, H, O, and a variety of fission products (FPs) were analyzed for
the equilibrium composition for element ratios simulating burnup
in reactor. In addition to the chemical states, Thermo-Calc also pre-
dicts the phases present at equilibrium. Oxygen was assumed to
be present as an impurity (in the form of ZrO
2
) at a concentra-
tion of 1000wt ppm. Unfortunately, the U/Zr and H/Zr ratios were
not specified, but they are taken here to be 0.31(45wt%) and 1.6,
respectively.
Table 4 summarizes the states of fission products in irradiated
UO
2
and (U
0.31
Zr)H
1.6
O
0.01
. The former are taken from Olander
(1976) and the latter from Huang et al. (2001).
4. In-reactor chemical behavior
4.1. Impurity oxygen
The small quantity of impurity oxygen in the fresh hydride fuel
is present as ZrO
2
. The oxides of the rare-earth elements and of Ba
and Sr are more stable than ZrO
2
, so as the fission products (FP)
3
Justification for this omission is based on the claim that the post-irradiation
anneal data inFig. 11applytofuel that has beenirradiatedlongenoughfor all fission-
product activity to achieve equilibrium, when (it is claimed), the release fraction is
a maximum (Chesworth and West, 1985). This does not obviate the need to report
the length of time of the anneal.
grow in, the following reaction takes place:
2RE + 1.5ZrO
2
+1.5xH → RE
2
O
3
+1.5ZrH
x
(19a)
2(Ba, Sr) + ZrO
2
+xH → 2(Ba, Sr)O + ZrH
x
(19b)
The “H” in these reactions is taken from the hydride in order
to combine with the “liberated” Zr, thus reducing the value of x.
The details of this process are given in Appendix B. The end result
is reduction of the initial H/Zr ratio of 1.6 to 1.59 when all oxygen
has been transferred from ZrO
2
to the fission-product oxides. For
fuel with a fission rate density of 2×10
13
cm
−3
s
−1
, this stage is
complete after approximately 4 months of irradiation.
4.2. Oxygen potential in hydride fuel
The oxygen potential (^
¯
C
O
2
= RT ln p
O
2
) in a fuel rod is
important because it determines (thermodynamically) whether the
cladding inner wall can be corroded, as it is in oxide fuel. The only
oxygenpresent inhydride fuel is animpurity, andit becomes bound
as rare-earth or alkali-earth oxides. Accordingly, the oxygen poten-
tial is expectedto be very low. Just howlowcanbe determinedfrom
the thermochemistry of the following reaction
4
(RE)
2
O
3
+2H
2
= 2(RE)H
2
+
3
2
O
2
(20)
Although not technically a rare earth, yttrium is very similar
in its chemistry to the true members of this group. Consequently,
Y will be taken as a surrogate for RE. The above reaction can be
broken into component formation reactions and their associated
free energy changes at 800K:
RE +H
2
= (RE)H
2
^C
f
(RE)H
2
= −26kcal/mole (21H)
2RE +
3
2
O
2
= (RE)
2
O
3
^C
f
(RE)
2
O
3
= −400kcal/mole (21O)
Subtracting (21O) from twice (21H) gives reaction (20), so the
standard free energy change of the latter is
^C
o
(20)
= 2^C
f
(RE)H
2
−^C
f
(RE)
2
O
3
= 348kcal/mole
and the law of mass action gives the oxygen pressure:
p
O
2
= [exp(−348 ×10
3
]1.986 ×800) ×p
2
H
2
]
2]3
= 2 ×10
−67
atm
or an oxygen potential of −245kcal/mole. Zirconium does not oxi-
dize at this oxygen pressure, so the cladding inner surface should
not become oxidized.
4
It is immaterial that thereactionas writtendoes not actuallyoccur; it is sufficient
that all species it contains are present in the system.
1416 D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424
4.3. Reduction of the H/Zr ratio during irradiation
After conversion of impurity ZrO
2
to the FP oxides, all subse-
quent REandBa,Sr producedbyfissionextracts hydrogenfromZrH
x
to formthe hydrides (RE)H
2
and (Ba,Sr)H
2
. The result is a continual
reduction in the H/Zr ratio of the hydride.
^N
H
=
2(Y
RE
+Y
Ba,Sr
)
1.5Y
RE
+Y
Ba,Sr
(2.6 ×10
−6
)(t
d
−190)
= 3.8 ×10
−6
(t
d
−190)
After 1 year of irradiation, the quantity of hydrogen bound in FP
hydrides is 6.7×10
−4
moles/cm
3
. The initial hydrogen concentra-
tion is (from the numbers in the previous paragraph).
9.4 ×10
−3
moles H/ghydride ×8.3g hydride/cm
3
= 7.8 ×10
−2
moles H/cm
3
So, a fraction 6.7×10
−4
/7.8×10
−2
=8.5×10
−3
of the hydrogen
present has been removed from the ZrH
1.59
, further reducing the
H/Zr ratio to 1.576. Every 6 months, ∼1% of the hydrogen in the
ZrH
x
matrix is transferred to fission products.
This process may not be as serious as it first appears. The hydro-
gen removed from the Zr remains in the fuel as FP hydrides, so
the total hydrogen concentration is unchanged. Since the fuel does
not lose hydrogen, the neutronics of the system are not affected by
the switch of the metal atoms to which the hydrogen is attached.
Moreover, The fission-product hydrides are more stable thermo-
chemically than ZrH
x
, so at a given temperature, the equilibrium
H
2
pressure over a mixture of ZrH
x
and FPH
2
is lower than that
over pure ZrH
x −y
, where y depends on the burnup, as indicated
above. The lower H
2
pressure reduces the risk of overpressurization
of the rodinthe event of a transient that substantially increases fuel
temperature.
4.4. Hydrogen loss to the gas phase
During normal operation, loss of hydrogen from ZrH
1.6
to the
gas phase in the fuel element is not significant, owing to the low
equilibrium H
2
pressure at fuel temperature. However in acci-
dent conditions accompanied by high temperature excursions and
cladding failure, rapid hydrogen release may constitute a unique
hazard. Release from the fuel occurs when the equilibrium hydro-
gen pressure of the fuel (from Fig. 6, modified by conversion to
fission-product hydrides) exceeds that in the gap or plenum gas.
The dehydriding reaction is:
ZrH
x
→ZrH
y
+
1
2
(x −y)H
2
(22)
In order to determine the kinetics of this reaction, a specimen
of ZrH
x
was suspended from a microbalance in a vacuum furnace
and the mass loss recorded as a function of time (Gutkowski, 2005).
The top graph in Fig. 12 shows a typical result. The most significant
aspect of this plot is its linearity. This is wholly unexpected from a
mechanistic point of view. One of two steps can control the kinetics
of this reaction. The first is diffusion of hydrogen (as interstitial H
atoms) in the zirconiumhydride matrix to the free surface. The sec-
ond is recombination of H atoms at the surface to form H
2
, which
then escapes to the gas phase. For either rate-limiting step, the rate
should have decreased with time, with a consequent upward cur-
vature of this plot. The constant dehydriding rate implied by the
straight line in the top graph implies zero order for the above reac-
tion (i.e., the rate is independent of Hconcentration in the hydride).
These experiments needtobe repeatedina typical in-reactor atmo-
sphere (∼20atm He) in order to verify this strange behavior.
Fig. 12. Dehydriding kinetics of ZrHx; top: mass loss (converted to H/Zr ratio) as a
function of time at 692

C; bottom: Arrhenius plot of rate data (Gutkowski, 2005).
The bottom graph in Fig. 12 is a conventional Arrhenius plot of
the dehydriding rates. The line is represented by
rate = 8 ×10
11
exp


34, 000
1

mmoles H
cm
2
-s
The number in the exponential term corresponds to an activa-
tion energy of 280±60kJ/mole.
The significance of these results can be appreciated from the
following example in a fuel rod with breached cladding. Suppose a
1-cm diameter, 1-cm long pellet of U
0.31
ZrH
1.6
is at a uniform tem-
perature of 700

C. The ∼80 volume percent hydride in this fuel
has a density of 5.64g/cm
3
. The pellet contains 0.0475moles of
Zr and 0.076moles of H. From Fig. 12, the H flux at this temper-
ature is ∼3×10
−7
moles H/cm
2
-s, so the rate of H loss from the
sides of the pellet exposed to the open gap is ∼10
−6
moles H/s.
In 1h, the pellet loses 3.6×10
−3
moles H, which corresponds to
a reduction in the H/Zr ratio from 1.6 to 1.52. If continued, this
rate of H loss may lead to a significant buildup of H
2
in the ex-
core environment and to increased reactivity of the H-depleted fuel
with steam. In intact LWR fuel rods, on the other hand, the small
plenum volume allows the H
2
pressure to build up to its equilib-
rium value before significant H depletion of the fuel occurs (see
Section 5).
4.5. Stability of hydride fuel in water
The compatibility of hydride fuel with coolant water is a prime
consideration in the use of this material in LWRs. Although long-
term tests of the fuel in high-pressure water at ∼300

C have not
been performed, quench tests of TRIGA fuel (identical in com-
position to the proposed power-reactor fuel) have been reported
(Lindgren and Simnad, 1979). Specimens about the size of a BWR
oxide fuel pellet were heated up to 1200

C (atmosphere not spec-
ified) for a matter of minutes and dropped into water. The authors
report the retrieved specimens to have suffered “minor cracking”
D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424 1417
Table 5
Operating temperatures of PWR fuel rods with hydride or oxide fuel. The fuel OD
is 12mm and the LHR is 37.5kW/m. The coolant temperature is 300

C. He and LM
gap-fillers are compared.
Temperature,

C U
0.31
ZrH
1.6
(He) U
0.31
ZrH
1.6
(LM) UO
2
(He)
Fuel centerline 680 555 1450
a
^T
fuel
170 170 665
^Tgap 125
b
1
+
125
b
^T
clad
46 46 46
T
fluid
39 39 39
a
UO
2
thermal conductivity=(22.7+7×10
−3
T(K)]
−1
.
b
For a gap thickness of 35␮m.
but otherwise were in excellent condition. The average weight loss
for 8 tests was 2%. This was attributed to hydrogen loss, partially
counterbalanced by surface oxidation.
5. Comparison of liquid-metal-bonded hydride- and
oxide-fueled LWRs
Estimated temperatures are shown in Table 5 for three PWR fuel
elements: two hydride fuels, one with helium in the fuel-cladding
gap and the other with a liquid-metal bond. For comparison, the
comparable temperatures in a conventional oxide-fueled system
are also given.
This table illustrates some of the favorable features of hydride
fuel relative to oxide fuel. The first is the much lower fuel tempera-
tures for the same LHR. This is due to the large thermal conductivity
of the hydride fuel, which is approximately equal to that of Zircaloy
(Zry). However, the important consideration is not the tempera-
tures per se; it is how the lower temperatures in the hydride affect
fuel performance. At the maximum temperature of the helium-
bonded hydride fuel, fission product release is due to recoil only
(see Section 3.4), whereas the UO
2
fuel is hot enough for release
of fission gases by diffusion to be significant. The ∼4-fold lower
^T
fuel
in the hydride fuel compared to the oxide version consider-
ably reduces the thermal stresses in the former, probably enough
to avoid pellet cracking entirely. For the conditions given in Table 5,
the stored energy in the helium-bonded hydride fuel is one-quarter
of that in the oxide fuel at the same LHR. The factor is even
larger if the hydride fuel has a liquid-metal in the fuel-cladding
gap.
The choice of the H/Zr atom ratio of 1.6 is a compromise
betweenincorporating the largest possible hydrogenconcentration
for neutron-moderation without generating too large a hydrogen
overpressure. At 700

C, for example, Fig. 6 gives p(H
2
) =0.1atm for
ZrH
1.6
. Hydrogen gas filling a 15-cm
3
plenum at this pressure con-
stitutes less than 0.1% of the hydrogen in a standard LWR-size pellet
of (U
0.31
Zr)H
1.6
, so there is no risk of hydrogen depletion of the fuel
by this mechanism. The choice of an H/Zr ratio of 1.6 places the fuel
squarely in the center of the ␦-Zr–H phase, which eliminates the
danger that modest composition variations will result in precipita-
tion of another phase. However, the H
2
pressure does not limit the
maximum operating temperature during normal operation.
Hydride fuel-element designs call for replacement of helium in
the fuel-cladding gap by a liquid metal (LM). The alloy of choice
is the ternary Pb–Sn–Bi eutectic, which melts at ∼100

C and is
chemically benign(Wongsawaeng andOlander, 2007). It serves two
important purposes. The first is to protect the cladding inner wall
fromnoxious fission products. The second is to reduce the fuel tem-
peratures by much100

Cor more comparedtohelium-bondedfuel
elements.
Demonstration of the thermal benefits of LM-bonded fuel
requires a test-reactor irradiation. Until this is done, the follow-
ing simple transient test served the same purpose. Helium- and
LM-bonded, UO
2
-fueled, Zircaloy-clad rodlets were quenched from
Fig. 13. Transient responses of fuel centerlinethermocouples in3-pellet rodlets with
helium or a liquid metal in the fuel-cladding gap.
600

C by immersion in a large body of water at 25

C. Each rodlet
had a thermocouple in a hole drilled into the centerline of the
middle pellet in order to measure the transient response of the
fuel centerline temperature. The external heat-transfer coefficient
was determined by performing the same test on a solid copper
rod of the same diameter as the miniature fuel elements. Fig. 13
shows the experimental results along with the theoretical predic-
tions obtained by solving the heat conduction equation in fuel with
the boundary condition accounting for the series resistances due
to the fuel-cladding bond material, the cladding, and the external
fluidfilm. Theonlydifferencebetweenthetworodlets was thebond
material.
Two conclusions can be drawn from this test. The first is that
the LMis a much more efficient heat transfer mediumthan helium,
which is not surprising considering the difference in the thermal
conductivities (k
LM
∼100k
He
). The second feature of Fig. 13 is the
good agreement between the calculated and experimental curves
for the LM bond but the vast discrepancy for helium in the gap.
The latter is due to the large radial gap thickness (estimated at
250␮m cold) and the likelihood of off-center pellets. However,
these geometrical details are unimportant for the LM-bonded ele-
ment, because the gap thermal conductivity is so large that there
is no temperature drop across it irrespective of its size or align-
ment.
6. Hydride fuel fabrication
A glance at the phase diagram in Fig. 6 suggests that fabri-
cating massive, crack-free pieces of (U
0.31
Zr)H
1.6
is likely to be a
costly process. Following preparation by arc melting the pure met-
als, the U–Zr alloy needs to be hydrided carefully. Adding hydrogen
at constant temperature moves the system through three phases:
first ␣-Zr, then the two-metal-phase region, then ␤-Zr, then the
metal-hydridetwophaseregionandfinallyintothe␦hydridesingle
phase. Volume changes in this process are significant. The zirco-
niumatomdensity in ␤-Zr is 16% greater than that in the ␦-hydride,
which implies a volume increase on hydriding of the same amount.
In addition, the crystal structure changes.
Fig. 14 shows the flowchart for producing (U,Zr)H
x
pellets ready
for insertion into cladding. Uraniumhexafluoride received fromthe
enrichment plant is reduced to the tetrafluoride with hydrogen gas,
leaving a waste of corrosive HF to be disposed of. Fig. 15 shows the
apparatus in which this step is performed. The conditions are an H
2
pressure of 1atm and a temperature of ∼500

C. The UF
4
powder
collected at the bottom of the reaction vessel is transferred to a
reduction furnace (called a bomb), in which Ca (or Mg) reduces the
1418 D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424
Fig. 14. Overall hydride-fuel pellet production flow diagram.
UF
4
to a metal:
UF
4
(s) + Ca → U(L) + CaF
2
(s)
This reaction is highly exothermic, so care must be taken to
prevent runaway temperatures. The final temperature is above the
meltingpoint of uranium(1132

C). The CaF
2
slagfloats ontopof the
liquid metal, so separation is relatively straightforward. The prod-
uct of this step is an ingot of impure uranium (mainly bits of slag
and unreacted calcium).
The alloy is made by melting the uranium ingot along with the
appropriate amount of zirconium metal in the induction furnace
shown in Fig. 16. In this apparatus, the remaining calciumvaporizes
and the CaF
2
separates as a slag. Once homogenized in the liquid
phase, the U–Zr alloyis releasedfromthe crucible andflows intothe
molds (called moules in the figure) which are roughly the diameter
of the fuel pellet.
After breaking the rough pellet-size ingots out of the molds, the
alloy surface is cleaned and the ingot transferred to the hydriding
furnace shown in Fig. 17. Thorough removal of oxides and nitrides
from the surface is essential because hydrogen absorption is eas-
ily impeded by this type of contamination. The cleaning procedure
includes:
1. degreasing in trichloroethylene;
2. abrading (machining, filing, grit blasting, grinding, wire brush-
ing);
3. pickling with a solution of HNO
3
and HF;
4. washing in distilled water, alcohol.
Addition of hydrogen to the alloy must be done very carefully
to avoid cracking the hydride because of the volume change men-
tioned above. The preferred method is to heat the metal to ∼800

C
and add hydrogen at ∼100Torr above the equilibrium pressure of
the ␤-Zr/␦ two-phase region, which is 130Torr at this temperature.
The objective is to bring the H/Zr ratio to about unity, which is the
terminal solubility of hydrogenin␤-Zr. As long as the hydride phase
is not formed, there is no risk of cracking the metal. The hydrogen
pressure is then increased to 1atm, and the gas is admitted in small
pulses to avoid too-rapid absorption in the metal, which would risk
cracking of the hydride being formed. Additional details of the fab-
rication process are provided by Huffine (1968) and in many of the
patents for TRIGA fuel elements (Eggers, 1978; Simnad et al., 1964).
7. Gap closure
The objective of this section is to determine the thickness of the
fuel-cladding gap as a function of burnup. The fuel is U
0.31
ZrH
1.6
and the gap is filled with a liquid metal. Because of the low fuel
temperature, there is no fission-product release (Fig. 11) and the
only fission-related swelling is due to solid and all volatile fission
products (Section 3.3).
7.1. Fuel swelling
The variables affecting this calculation are the initial fuel radius
R
Fo
and the fuel centerline and surface temperatures, T
o
and T
S
.
These are taken from Table 5.
The increase in the fuel radius, ^R
F
, is related to the azimuthal
strain at the fuel surface, ε
tot

, by
^R
F
= ε
tot

×R
Fo
(23)
where
ε
tot

= ε
th


H


FP

(24)
The three terms on the right are due to the following expansion
mechanisms: temperature distribution; hydrogen redistribution
and swelling due to solid fission products and probably xenon and
kryptonas well. The first twoterms are calculatedusing the method
by Olander and Ng (2005).
7.1.1. Thermal expansion
The conventional method for calculating ε
th

consists of sum-
ming the unrestrained expansion of differential rings in the pellet:
ε
th

=
2
R
2
o

Ro
0
˛(1) ×(1 −1
ref
)r dr (25)
If the thermal expansion coefficient ˛ is independent of temper-
ature and the temperature distribution is parabolic, Eq. (25) gives
the well-known result: ε
th

= ˛[1]2(1
o
+1
S
) −1
ref
]where T
ref
is the
cold temperature (25

C). In the case of a temperature-dependent
linear thermal expansion coefficient, however, this method gives
incorrect results; thermoelasticity theory must be used.
Following the procedure used in Appendix A of Olander and Ng
(2005), stresses are expressed in dimensionless form:
S
th
i
=
o
th
i
o

th
(26)
where i =r, 0 or z and:
o

th
=
˛
o
£(1
o
−1
S
)
1 −¡
(27)
The physical properties of the fuel are given in Table 3:
Young’s modulus is E=130GPa and Poisson’s ratio is ¡ =0.32.
D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424 1419
Fig. 15. Apparatus for reducing UF
6
(g) to UF
4
(s) (Sauteron, 1965).
The thermal expansion coefficient of ZrH
x
is temperature-
dependent:
˛ = ˛
o
(1 +u1) (28)
where ˛
o
=7.4×10
−6 ◦
C
−1
and a =2×10
−3 ◦
C
−1
.
From Table 6, T
o
−T
S
=170

C. With these parameter values, the
reference thermal stress of Eq. (27) is: o

th
= 241MPa.
The radial strain due to thermal expansion is
ε
th

=


th
£
[S
th
0
−¡(S
th
r
+S
th
z
)] +˛(1 −1
ref
) (29)
The dimensionless stresses at the fuel surface are (Olander and
Ng, 2005):
S
th
r
= 0
S
th
0
= S
th
z
= (1]2)(1 +u1
S
) +(1]3)u(1
o
−1
S
)
(30)
Substituting Eq. (28) (at the surface) and Eq. (30) into Eq. (29)
yields:
1
˛
o
ε
th

= (1
o
−1
S
)[(1]2) +(1]3)u(1
o
+1]21
S
)] +(1 +u1
S
)(1
S
−1
ref
)
(31)
1420 D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424
Fig. 16. Induction-melting furnace for production of U–Zr ingots (Sauteron, 1965).
where Eq. (26) has been used to dimensionlize the stress compo-
nents.
In terms of t
o
=T
o
−T
ref
and t
S
=T
S
−T
ref
, Eq. (31) is
1
˛
o
ε
th

= (t
o
−t
S
)[(1]2) +(1]3)u(t
o
+(1]2)t
S
+(3]2)1
ref
)]
+t
S
+u ×t
S
(t
S
+1
ref
) (32)
The “1/2” in the brackets and “t
S
” following it combine to give
the constant-␣ result. The coefficient of “a” is
coef. of a = (1]2)1
ref
(t
o
+t
S
) +(1]3)(t
2
o
−(1]2)t
o
t
S
+(5]2)t
2
S
) (33)
The conditions given in Table 5 are used: T
o
=555

C, T
S
=385

C.
In addition, T
ref
=25

C, ˛
o
=7.4×10
−6 ◦
C
−1
, a =2×10
−3 ◦
C
−1
.
The thermal straincorrespondingto␣
o
alone is 3.3×10
−3
. Using
Eq. (33), the contribution from the temperature-dependent part of
˛ is 2.7×10
−3
. The total thermal strain at the pellet surface is
ε
th

= 6.0 ×10
−3
(34)
7.1.2. Hydrogen expansion
The strain arising from volume changes due to hydrogen redis-
tribution is given by
ε
H

=
o

H
£
[S
H

−¡(S
H
r
+S
H
z
)] +ˇ(x −1.6) (35)
where x is the H/Zr ratio and ˇ=0.027 (Table 3). The reference
hydrogen stress is
o

H
=
ˇ£
1 −¡
(36)
With the given parameter values, o

H
= 5200MPa. Fromthe formu-
las given in Section 6 of Olander and Ng (2005), the dimensionless
stress components at the surface of the fuel due to hydrogen redis-
tribution are
S
H
r
= 0 S
H
0
= S
H
z
= −0.174 (37)
The negative sign indicates that the surface stress components are
compressive.Substituting these values, along withH/Zr =1.78 at the
fuel surface(fromFig. 5), intoEq. (35) yields thefollowingazimuthal
strain due to this mechanism:
ε
H

= −4.72 ×10
−3
+4.86 ×10
−3
= 1.4 ×10
−4
(38)
The magnitude of the hydrogen stress contribution (first term in
Eq. (38)) is practically equal in magnitude to the component due to
unrestrained hydrogen expansion (second term in Eq. (38)). How-
ever, the two terms are of opposite sign, which renders this strain
component very small.
7.1.3. Solid fission-product swelling
As showninSection3.3, hydride fuel swelling due tosolidfission
products is proportional to burnup:
ε
FP

=
1
3

^v
v

FP
=
1
3
×2 ×10
−3
×BU
equiv. ox
(39)
7.1.4. Total fuel-surface strain
Substituting Eqs. (34), (38) and (39) into Eq. (24) gives the total
azimuthal strain at the pellet surface:
ε
tot
F
= 6.0 ×10
−3
+1.4 ×10
−4
+6.7 ×10
−4
BU
equiv. ox
(40)
7.2. Cladding creepdown
Because of the pressure difference across the cladding, creep
causes a slow reduction in the cladding radius. This phenomenon
is especially severe in PWRs because of the thinner cladding and
higher operating temperature compared to those in BWRs. The
model used for analyzing this process is that proposed by Limback
and Andersson (1996), which is also used in the FALCON fuel-
performance code (Lyon et al., 2004). In the following equations of
the model, numerical values of the parameters for stress-relieved-
annealed (SRA) Zry-4 are used.
D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424 1421
Fig. 17. Hydriding furnace (Van Houten, 1974).
7.2.1. Secondary thermal creep
The steady-state secondary strain rate is
˙ ε
ss
= 1.08 ×10
9
£
1

sinh
u
i
o
0
£

2
exp


24, 200
1

(41)
T is the temperature in K and o
0
is the azimuthal stress in MPa. The
creep strain rate is in h
−1
.
a
i
accounts for irradiation hardening of thermal creep, which
results from the accumulation of immobile dislocation loops and
point defects:
u
i
= 650[1 −0.56{1 −exp(−1.4 ×10
−27
˚
1.3
)}] (42)
The hardening effect depends on the fast fluence, ˚, in n/cm
2
. ˙ ␧
ss
is thus a known function of time.
7.2.2. Primary thermal creep
The strain due to primary thermal creep is given by:
ε
p
= ε
sat
p
[1 −exp(−52

˙ ␧
ss
t)] (43)
where t is the time in hours and ε
sat
p
is the saturation value of
primary creep:
ε
sat
p
= 0.0216 × ˙ ␧
0.109
ss
[2 −tanh(35, 500 × ˙ ␧
ss
)]
−2.05
(44)
The time variation of ˙ ␧
ss
is taken into account in calculating
primary creep in the following approximate manner. Over a time
interval ^t, ˙ ␧
ss
in Eqs. (43) and (44) is assumed to be constant at
the average value in the interval. Only the explicit time dependence
in Eq. (43) is retained. The result is:
ε
p
(t +^t) = ε
p
(t) +ε
sat
p
(e
−52

˙ ␧ss

t
−c
−52

˙ ␧ss

t+^t
) (45)
In this equation, ˙ ␧
ss
and ␧
sat
p
are evaluated at t +1/2^t.
7.2.3. Irradiation creep
The irradiation creep strain depends on both stress and the fast
flux ¢(n/m
2
-s), but time t is in hours
5
:
ε
irr
= 3.56 ×10
−24
¢
0.85
o

×t (46)
7.2.4. Total cladding creep strain
The total cladding creep strain as a function of time (in hours) is
given by
ε
tot
C
= ε
p

ss

irr
(47)
The creep strains in this equation are considered to be positive
quantities even though they apply to creepdown of the cladding
tube. This is taken into account in Eq. (49).
7.3. Gap closure
7.3.1. Dimension changes
The total creep strain of Eq. (47) is applied to the mean radius of
the cladding, the initial value of which is
¯
R
o
C
=
1
2
(D
O
−ı
C
) (48)
D
O
is the cladding outer diameter and ı
C
is the cladding thickness.
At time t, the mean cladding radius is
¯
R
C
=
¯
R
o
C
(1 −ε
tot
C
) (49)
with the strain taken from Eq. (47). The inside wall radius is
R
Ci
=
¯
R
C

1
2
ı
C
(50)
The initial fuel radius (cold) is
R
o
F
=
1
2
D
O
−ı
C
−ı
o
g
(51)
where ı
o
g
is the thickness of the initial cold gap.The radius of the
fuel pellet is
R
F
= R
o
F
(1 +ε
tot
F
) (52)
where ε
tot
F
is given by Eq. (40).
7.3.2. Parameters of typical hydride fuel element
The following parameter values are utilized to apply the model
to a typical hydride fuel element in a BWR:
Geometry (in mm):
D
O
=12
ı
C
=0.78
ı
o
g
= 0.08 and 0.15
Thermomechanical state:
5
The creep model of Limback and Andersson (1996) unfortunately mixes units.
The fast flux in Eq. (46) is in n/m
2
-s, while the fast fluence in Eq. (42) is in n/cm
2
.
The two are related by ˚=¢t. Strain rates are expressed in h
−1
but the flux utilizes
seconds.
1422 D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424
^p=pressure difference betweencoolant androdinterior =3MPa.
The hoop stress (compressive) in the cladding is:
o
0
= ^p
1/2D
O
ı
C
= 3 ×
6
0.78
= 23MPa
T =meancladdingtemperature. For the thermal conditions inTable
6, T =635K.
E=Young’s modulus for SRA-Zry-4. At 635K it is 7.7×10
4
MPa
(Limback and Andersson, 1996).
Irradiation conditions:
¢=10
18
n/m
2
-s
The fast fluence is:
˚ = (¢ ×10
−4
)(t ×3600) n/cm
2
˙
F = 3 ×10
13
fissions/cm
3
-s
t =time, h
Steady-state creep rate:
For the conditions listed above, the steady-state creep rate given
by Eqs. (41) and (42) is fitted to the following formula:
˙ ␧
ss
= (0.242 +1.22e
−0.001×t
) ×10
−7
, h
−1
time and burnup are related by
BU
equiv.ox
= 3.2 ×10
−11
J
fiss
×3600
˙
Ft
fiss
cm
3
×10
−6
MW
W
×
10
3
¡
U
cm
3
kgU
×
1
3600 ×24
d
s
×
¡
U
9.3
The last ratio converts burnup to equivalent-oxide burnup. With
the fission density given above, the conversion is:
BU
equiv. ox
= 7.2 ×10
−3
t
7.3.3. Results
Fig. 18 shows that the three components of the cladding creep-
down strain are of comparable magnitude, although irradiation
creep starts to dominate after ∼25MWd/kgU equivalent-oxide
burnup.
Fig. 19 shows the variations of the fuel and cladding ID radii
as functions of burnup. Gap closure occurs when the fuel and
cladding lines intersect. The fuel radii at time zero include the
Fig. 18. Components of cladding creepdown.
Fig. 19. Gap closure—fuel and cladding inner radii as functions of burnup.
strains due to thermal and hydrogen stresses. The difference in
the initial fuel radii for the two cold gap thicknesses is due to
the requirement that the cladding geometry remain constant. The
increase thereafter is due solely to fission-product swelling. With
an 80␮m initial cold gap, PCMI occurs at an equivalent-oxide
burnup of ∼10MWd/kgU. With a 150-␮m initial gap, closure is
delayed until ∼27MWd/kgU equivalent-oxide burnup. The goal
of avoiding gap closure until 60MWd/kgU cannot be achieved
using the liquid-metal in the gap unless the initial gap is ∼300␮m.
According to Eq. (51), a gap of this size contains fuel of 4.92mm
radius, compared to 5.17mm for the 80-␮m cold radial gap. For a
fixed linear heat rate, the enrichment varies as the square of the
fuel pellet diameter. In order to maintain the same LHR for the two
fuel pellet dimensions, the enrichment of the small pellet must be
9% larger than for the large pellet. If fuel the in the rod with the
80-␮m gap is has an enrichment of 12%, this figure needs to be
increased to 13% if the gap thickness is increased to 300␮m.
The reason for the necessity of a large initial gap to avoid PCMI
at ∼60MWd/kgUis the sizeable fission-product swelling rate of the
hydride fuel. According to the only data available, this quantity is
three times larger in hydride fuel than in oxide fuel.
8. Conclusions
Building on the early work on hydride reactor fuel in the SNAP
and TRIGA programs, an assessment of the potential for its use
as LWR fuel is presented. The principal properties of (U,Zr)H
x
,
notably its high thermal conductivity, provide this material with
several attractive features as a nuclear fuel. The low fuel temper-
ature reduces fission-product release to negligible amounts and
decreases storedenergyinthe fuel. Witha low-melting liquidmetal
inthe fuel-cladding gap, PCMI canbe delayedor avoidedby increas-
ing the thickness of the cold gap. Initial fuel cracking due to thermal
stress may be avoidedbecause of the countervailing effect of hydro-
gen redistribution.
A cautionary note is the appearance of voids produced by
fission-fragment damage the hydride lattice adjacent to the ura-
nium particles. The swelling (^V/V) due to the voids appears early
in irradiation (burnup ∼10
−3
FIMA) and is highly temperature-
sensitive: it is absent at 650

C, reaches 1%at 700

Cand5%at 750

C.
Liquid-metal bonding of the fuel-cladding gap is necessary to avoid
temperatures where severe initial swelling occurs.
Although many of the essential radiation properties of hydride
fuel have been determined, a well-instrumented irradiation test
D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424 1423
under power-reactor conditions is needed, particularly in order to
assess (i) the effect of released fission gas (if any) on the thermal
effectiveness of the liquid metal bond in the fuel-cladding gap;
(ii) whether the initial void swelling occurs during power-reactor
conditions. Additional laboratory tests include measurement of
hydrogenloss as afunctionof temperaturein∼20atmof heliumand
assessing the long-termcompatibility of the hydride fuel with PWR
coolant water at 300

Candwithsteamat T >1000

Crepresentative
of accident conditions.
A significant unknown is the cost of large-scale fabrication of
hydride fuel elements, which, when added to the increased enrich-
ment cost, may compromise its competitiveness vis-à-vis oxide
fuel.
Avoiding closure of the fuel-cladding gap during irradiation
may be problematic because of the possibly large fission-product
swelling rate of hydride fuel compared to oxide fuel. In order to
overcome this feature (if it is true), the thickness of the LM-filled
fuel-cladding gapmust be made verylarge (∼300␮mradial), which
requires a ∼1% enrichment increase to maintain the design LHR.
Appendix A. Permeation of hydrogen through stainless
steel
Dissolution is accompanied by dissociation of H
2
into atoms, so
the equilibrium between gaseous hydrogen and the metal:
H
2
(g) = 2H(dissolved) (A1)
is given by Sieverts’ law:
C
H
= K
H

p
H
2
(A2)
where C
H
is the concentration of H in the metal in moles/m
3
, p
H
2
is the hydrogen pressure in Pa and K
H
is the Sieverts’ law constant.
The flux of H atoms through stainless steel of thickness ı
SS
with
H
2
pressure p
H
2
in the fuel-cladding gap and zero pressure at the
Zry/SS interface is:
j
H
= D
H
C
H
ı
SS
= (D
H
K
H
)

p
H
2
ı
SS
(A3)
The product D
H
K
H
is the permeability of hydrogen in the metal. For
stainless steel, LeClaire gives the equation (LeClaire, 1984):
D
H
K
H
= 3.3 ×10
−7
exp


7900
1

moles/m-s-Pa
0.5
(A4)
To assess the utility of this protection scheme in the present
design, consider a 0.4-mm-thick SS liner at 340

C exposed at its
inner surface to the equilibrium hydrogen pressure generated by
the hydride fuel (∼5×10
−3
atm, Fig. 5). The above two equations
give a permeation rate J
H
=5×10
−12
moles H/cm
2
-s. The fractional
change rate of the H/Zr ratio (x) of the fuel is
d(x/x
0
)
dt
= −
j
H
N
H0
d
(A5)
where x
0
is the initial H/Zr ratio (1.6), N
H0
=0.08moles H/cm
3
is the
molar density of hydrogen in ZrH
1.6
and d is the ID of the cladding
(∼1cm).
Appendix B. Effect of fission products on H/Zr ratio of fuel
Reaction of impurity oxygen
During the early stage of irradiation, zirconium attached to
oxygen impurity as ZrO
2
is released as the oxygen is transferred
to fission products by reactions (19a) and (19b). The zirconium
released by these reactions, along with fission-product, Zr join that
in the original ZrH
1.6
. Hydrogen is not affected by this conversion.
Fig. B(a) depicts this process using a basis of 1 mole of ZrH
1.6
and N
O
Fig. B1. Changes in H/Zr ratio during irradiation. (a) Change of H/Zr ratio during
transfer of oxygen from Zr to Re and Ba,Sr and (b) Reduction in H/Zr ratio with
burnup.
moles of oxygen. In order to move this quantity of oxygen fromZr to
the reactive fission products, n moles of U are fissioned per mole of
initial Zr. During this period, nY
Zr
moles of Zr are produced, where
Y
Zr
is the fission yield of zirconium. The H/Zr ratio at this point is:
H
Zr
=
1.6
1 +nY
Zr
+1]2N
O
(B1)
The moles fissioned required to consume the N
O
moles of impu-
rity oxygen as the oxides in reactions (19a) and (19b) is obtained
from:
N
O
= 3]2N
RE
+N
Ba,Sr
= (3]2Y
RE
+Y
Ba,Sr
) ×n
or, solving for n:
n =
N
O
3]2Y
RE
+Y
Ba,Sr
(B2)
The amount of oxygen corresponding to ∼1000ppmwt in the
fuel is N
O
∼0.011moles O per mole Zr. Using this value, along with
the yields from Table 4, Eq. (B2) gives n=0.012moles fissioned per
mole of initial Zr. Substituting these values of N
O
and n into Eq. (B1)
gives H/Zr =1.586 at the point of complete reaction of ZrO
2
.
Using the information in Table 1, the molar density of zirconium
in (U
0.31
Zr)H
1.6
is:
¡
Zr
= 8.0
g fuel
cm
3
×0.55
g Zr
g cm
3
×
1 mole
91.2 g Zr
= 0.050
moles Zr
cm
3
(B3)
The fission density required for this process is:
F∗ = n ס
Zr
×N
Av
= 1.2 ×10
−2
×0.050 ×6.02 ×10
23
= 3.6 ×10
20
fissions]cm
3
where N
Av
is Avogadro’s number. Assuming a fission rate of
2×10
13
fissions/cm
3
-s, thetimetoachievetheabovefissiondensity
is 1.8×10
7
s, or about 7 months.
Using Eq. (13), F* can be converted to burnup:
F ∗ ]N
Av
= ¡
Zr
(1 +y) ×FIMA ∗ moles fissioned]cm
3
Substituting the above values of F* and ¡
Zr
along with y =0.31
yields FIMA* =9.2×10
−3
.
Subsequent change in H/Zr with burnup
1424 D. Olander et al. / Nuclear Engineering and Design 239 (2009) 1406–1424
This step, shown in Fig. B(b), is based on one mole of ZrH
1.59
at
the burnup FIMA* of oxygen removal fromZrO
2
. Following n

addi-
tional fissions per mole of initial Zr, 1+n

Y
Zr
moles of Zr are present.
Since the active fission products form dihydrides, the quantity of
hydrogen is reduced to 1.59−2n

(Y
RE
+Y
Ba,Sr
). The H/Zr ratio is:
H
Zr
=
1.59 −2n

(Y
RE
+Y
Ba,Sr
)
1 +n

Y
Zr
(B4)
The number of moles of fissions is given by Eq. (13):
n

= (F −F∗)]N
Av
= ¡
Zr
(1 +y)(FIMA −FIMA∗)
= 0.066(FIMA −FIMA∗) (B5)
with ¡
Zr
given by Eq. (B3) and y =0.31.
The function obtained by substitution of Eq. (B5) into (B4) is
adequately represented by a line with a slope:
z(H]Zr)
z(FIMA)

= 0.1 (B6)
The relation between U-235 consumption and FIMA is obtained
from Eq. (13):
zN
235
N
0
235
=
(N
0
235
+N
0
Zr
) ×FIMA
c ×N
0
235
=
1 +y
−1
c
×FIMA = 33.8 ×FIMA
(B7)
The numerical coefficient in the above equation corresponds to
y =0.31 and e =0.125.
Time is related to FIMAaccording to Eq. (13), which, for constant
fission rate, yields:
˙
Ft = N
0
U
(1 +y
−1
) ×FIMA (B8)
For the uranium density given in Table 1 (3.7gU/cm
3
), N
0
U
=
9.3 ×10
21
atoms U/cm
3
. With a fission rate of 2×10
13
cm
−3
s
−1
,
Eq. (B8) becomes:
t = 774 ×FIMA mos (B9)
References
Albrecht, W., Goode, W., 1960. Diffusion of hydrogen in zirconium hydride. USAEC
Rept. BMI-1426.
Beck, R., Mueller, W., 1968. Zirconiumand HafniumHydrides. In: Mueller, W., Black-
ledge, J., Libowitz, G. (Eds.), Metal Hydrides, Section 7.3. Academic Press.
Chesworth, R., West, G., 1985, October. Final results of qualification testing of
TRIGA fuel. Presented at the International Meeting On Reduced Enrichment for
Research & Test Reactors, Petten, The Netherlands, October.
Eggers, G., 1978. Method of making ZrH fuel element. U.S. patent No. 4,071,587.
GA Technologies, Inc., unknown. Uranium-zirconium hydride TRIGA LEU fuel. In:
Research Reactor Core-conversion Guidebook, vol. 4, Fuels. IAEA-TECDOC-643,
Appendix I-7 (19).
Gelezunas, V., et al., 1963. The Diffusion Coefficients for Hydrogen in ␤-Zirconium.
J. Electrochem. Soc. 110, 799.
Greenspan, E., 1998. Personal Communication from Massoud Simnad.
Gutkowski, D., 2005. Kinetics of hydrogen uptake and release from zirconium
hydride. MS thesis. Dept. of Nuclear Engineering, University of California at
Berkeley.
Gylfe, J.D., et al., 1960. Evaluation of zirconium hydride as moderator in integral
boiling—water superheat reactors. NAA-SR-5943.
Hesketh, K.W., Stanbridge, J.R., 1993. Technology objectives for nuclear fuel in 2020.
In: Proceedings of GLOBAL-93, Seattle, WA, September.
Huang, J., et al., 2000. Hydrogen redistribution in zirconium hydride under a tem-
perature gradient. J. Nucl. Sci. Technol. 37, 887.
Huang, J., et al., 2001. Thermodynamic analysis of the chemical states of fission
products in U–Zr hydride fuel. J. Nucl. Mater. 294, 154.
Huffine, C., 1968. Fabrication of hydrides. In: Mueller, W., Blackledge, J., Libowitz, G.
(Eds.), Metal Hydrides, Chapter 13. Academic Press.
Iorgulis, C., et al., 1998. Neutronic calculations regarding the New LEU 6×6 fuel
bundle for the 14MWTRIGA. In: International Meeting on Reduced Enrichment
for Research Reactors, Sao Paulo, Brazil.
LeClaire, A., 1984. Permeationof hydrogenisotopes instructural alloys. J. Nucl. Mater.
122–123, 1558.
A. Lillie, et al., 1973. Zirconium hydride fuel element performance characteristics.
Atomics International/USAEC Rept. AI-AEC-13084.
Limback, M., Andersson, T., 1996. A model for analysis of the effect of final annealing
on the in- and out-of-reactor creep behavior of zircaloy cladding. In: Zir-
conium in the Nuclear Industry: Eleventh Symposium, ASTM STP 1295, pp.
448–468.
Lindgren, J., Simnad, M., 1979. LEU TRIGA fuel water-quench safety tests. Trans. Am.
Nucl. Soc..
Lyon, W.F., Jahingir, N., Montgomery, R.O., Yagnik, S., 2004. Capabilities of the FAL-
CON steady state and transient fuel performance code. In: Proceedings of the
2004 International Meeting on LWR Fuel Performance, Orlando, FL, September
19–22.
Majer, G., et al., 1994. J. Phys. Condens. Matter 6, 2935.
Merten, U., et al., 1958. Preparation and properties of U–Zr hydrogen alloys. In: Pro-
ceedings of the Second Conference on Peaceful Uses of Atomic Energy, P/789,
Geneva.
Olander, D.R., 1976. Nuclear reactor fuel elements. Nat’l. Tech. Info. Services. Docu-
ment No. 26711.
Olander, D., Ng, M., 2005. Hydride fuel behavior in LWRs. J. Nucl. Mater. 346, 98.
Olander, D.R., Wongsawaeng, D., 2006. Re-solution of fission gas—a review: Part I.
Intragranular bubbles. J. Nucl. Mater. 354, 94.
Paetz, V., Lücke, K., 1971. Z. Metallkund. 62, 657.
Parfenov, V., et al., 1967. Corrosion of zirconium and its alloys. Atomizdat. p. 179.
Sauteron, J., 1965. Les Combustibles Nucléaire. Herman.
Shuffler, C., Malen, J., Diller, P., Ganda, F., Todreas, N. Greenspan E., Petrovic, B., this
issue. Economic analysis for PWRs. Nucl. Eng. Des.
Shuffler, C., Malen, J., Trant, J., Todreas, N., this issue. Thermal hydraulic analysis for
grid-supported and inverted fueled PWR cores. Nucl. Eng. Des.
Simnad, M.T., 1981. The U–ZrHx alloy: its properties and use in TRIGA fuel. Nucl. Eng.
Des. 64, 403.
Simnad, M., et al., 1964. Method of hydriding. U.S. Patent No. 3,135,697.
Sommer, A., Dennison, W.,1960. Thermal diffusion of hydrogen in nonstoichiometric
zirconiuim hydride. USAEC Rept. NAA-SR-5066.
Toma, C., et al., 2002. Characterization of TRIGA LEU fuel behavior in 14 MW core.
Un-numbered report. Inst. For Nucl. Res., Romania.
V’yugov, P., Ryabchikov, L., 1979. Diffusionandsolubility of hydrogenin␤-zirconium.
Izv. Akad. Nauk. SSR Met. (Russian Metallurgy) 3, 214.
Van Houten, R., 1974. Engineering and fabrication aspects of nuclear metal hydrides.
Nucl. Eng. Des. 31, 434.
Wang, Wei-E., Olander, D., 1995. Thermodynamics of the Zr–Hsystem. J. Am. Ceram.
Soc. 78, 3323.
West, G., Simnad, M., Copeland, G., 1986. Final results from TRIGA LEU fuel postir-
radiation examination and evaluation following long-termirradiation testing in
the ORR. General Atomics report GA-A18641.
Wongsawaeng, D., Olander, D., 2007. Liquid-metal bond for LWR fuel rods. Nucl.
Technol. 159 (2007), 279.
Yamanaka, S., et al., 1999. Thermal and mechanical properties of zirconiumhydride.
J. Alloys Compd. 293–295, 23.
Yamanaka, S., et al., 2001. Thermal properties of zirconium hydride. J. Nucl. Mater.
294, 94.
Zuzek, E., Abriata, J., 1990. The hydrogen zirconium system. Bull. Alloy Phase Dia-
grams 11, 385.

Sign up to vote on this title
UsefulNot useful