Behaviour of hydrogen in

zirconium intermetallics
Patrick A. Burr
Sunday 8
th
May, 2011
1
Introduction
1.1 zirconium in industry
Zirconium (Zr) is seldom used in industry in its metallic form, and even then it is mostly
an alloying agent due to its good corrosion resistance. In other words, it is not a very
important metal for most applications. However, there is one niche market in which it
covers a fundamental role: the nuclear power generation industry. In fact, over 90% of
the Zr that is produced every year is for that market
1
. Zr is the cladding material of
choice for water cooled reactors (of all varieties: PWR, BWR, CANDU, VVER, RBMK, etc),
and it is also used as a structural material in some designs. Water cooled reactors are
by far the most widespread type of nuclear reactors in the world, with the Pressurised
Water Reactor (PWR) being the most common subtype. They operate with sintered fuel
pellets of Uranium dioxide usually enriched up to 5.5% of the ssile isotope with
dimensions of 1015 mm nominal diameter and 23 cm in height. The power produced by
the nuclear reaction is extracted in the form of heat by a ow of pressurised water around
the fuel. The role of the zirconium cladding (or sheath) is to mutually protect the fuel and
water from each other. Contamination of ssion products into the water stream would
inevitably increase the amount of radioactivity released into the biosphere, and the water
would corrode and erode the fuel pellets reducing their integrity, which in turn would
increase the chance of dispersion of both ssile elements and radionuclides. In addition
to encapsulating and protecting the fuel, the cladding also provides an easy means of
handling the many thousands fuel pellets by stacking them into long ( 4m) tubes, which
1
are then grouped into assemblies which are easy to deal with. In some reactors such as
the CANDU and the RBMK designs Zr alloys are adopted also for the pressure tubes
(and the calandria in the case of the Canadian model). The mechanical properties required
for such applications are radically dierent from that of the cladding, therefore dierent
alloys are engineered for this purpose.
The preferential choice of Zr as a cladding material over other, possibly cheaper, corrosion-
resistant structural materials such as steel, is a result of its exceptionally low cross-section
for neutron absorption (
a
= 0.185b,
a
= 0.00794cm
1
for thermal neutrons)
2
, and, more
importantly, the excellent cross-section to strength ratio. This implies that Zr acts as an
eective cladding material in that it retains its structural properties without poisoning
the neutron ux, which would otherwise have to be compensated by either higher fuel
enrichment, or better moderators; both of which would aggravate the economics of power
generation. When comparing the strength to cross-section ratio of the candidate materials
for cladding, it transpires that best options are Zr, Be and Mg (followed by a large margin by
Fe, Al and Ni). Magnesium, usedin Magnox reactors, was discarded for water reactors due
to the low melting temperature and more importantly high reactivity with oxygen.
Berylliumwas abandoneddue to high handling complexity andincreasedcosts thta arise as
a consequence of its toxicity in the powder form (though in more recent times the research
into Gen IV reactors has invigorated the attention to Be, together with more exotic cladding
materials, in order to overcome some of the intrinsic limitations of Zr).
Since the birth of water cooled reactors, there has been a large volume of research into
improving the mechanical and chemical properties of zirconium. There have been two
main paths for the engineering of zirconium alloys: one based on tin (Sn) as the main
alloying component, and the other based on niobium (Nb) additions. As we can see from
table 1.1, all Zircaloy claddings (developed by Westinghouse) fall under the former class.
Zircaloy 2 had much improved corrosion resistance compared to the initial alloy, however
nickel was later removed to form Zircaloy 4 as it is believed to contribute to the formation
of hydrides in the zirconium matrix
3,4
. At the same time many Nb-based zirconium alloys
have been developed all around the world; for example: the Russian Zr-1%Nb, E110, and
E635; the Canadian Zr-2.5%Nb; and the European Scanuk. Moreover, newer alloys such
as Zirlo and M5 have evolved taking advantage of Nb additions together with Sn or Fe,
however their exact composition is not available in the public domain.
1.2 hydrogen and Hydrides in zirconium
One of the main problems with zirconium cladding is hydrogen embrittlement. H in solid
solution sits in the many interstitial sites of the phase, described in Fig. ??, however it is
now widely accepted that H in solution does not have a great impact on the mechanical
2
Table 1.1: chemical compositions of commercial zirconium alloys. Values in
parentheses express electro-negativity (in Pauling scale) of each of the elements
taken in the pure state (from the CRC Handobook
5
). All other units are in wt%
unless otherwise stated.
Zr-Alloy (1.33) Sn (1.96) Fe (1.83) Cr (1.66) Ni (1.91) Nb (1.60) O (3.44)
Zircaloy 1
6
2.50 0.14 max
Zircaloy 2
4,610
0.100.05 1.450.25 0.1350.07 0.0550.025 0.14 max
Zircaloy 3 0.25 0.25 0.14 max
Zircaloy 4
4,611
1.450.25 0.210.03 0.10.03 0.14 max
Zr-1%Nb 0.61.0 0.090.13
Zr-2.5%Nb
?
< 0.13 2.42.8 0.090.13
Zirlo
6,12
0.71.5 0.070.14 { 0.03 0.14 } 0.52.0
M5
13?
0.014-0.037 1.00 0.12-0.15
properties of cladding, but rather that:
i. the mechanism through which hydrogen embrittles zirconium is by formation of
hydride platelets, which cause Delayed Hydride Cracking (DHC),
14
ii. the degree of embrittlement is strongly dependent on morphology and orientation of
the hydrides.
15,16
The main process for hydride formation is thermal cycling: at high temperatures
(T > 550

C) the solubility of H in Zr is very high (up to 5.9 at% in the phase

17
and
up to 50 at% in the ), and at operating temperature of roughly 300

C it is still between
0.40.7 at% (60 ppm in mass). On the other hand, the room temperature solubility is
completely negligible (see table 1.2 and Fig. 1.1 for further details).
Table 1.2: H solubility in -Zr at T=300

C
Paper at %
Kearns (1967)
9,15
0.7
Zuzek (1990)
17
< 1.0
Wang (1995)
21
< 1.0
Setoyama (2004)
20
0.47
_
52 2wt ppm
_
Okamoto (2006)
18
0.4 0.5
The direct implication is that during operation the hydrogen absorbed goes into solid
solution in the zirconium matrix, but if the cladding is allowed to cool down the hydrogen
3
Figure 1.1: The Zr-H pseudo-phase diagram from Okamoto
18
, with a
comparison of the data from Zuzek
17
, Dupin
19
and Setoyama
20
(left) and a
detail of the -Zr solubility of H (right).
content will be above the solubility limit and therefore will segregate in one of the , ,
(metastable) hydride forms. In normal operating conditions the thermal cycling is kept
at minimal levels (though unavoidable during re-fuelling and shuing of assemblies).
However, this becomes of particular interest in Reactivity Initiated Accidents (RIA), and
in dry storage environment, where a thermal spike is observed coupled with a potential
increase in hoop stress. The three hydride phases are described in the Fig. 1.2.
The hydrides grow in long chains of small, hard and very sharp lenticular particles. For
this reason the orientation of hydrides is crucial in the structural integrity of the cladding
tubes. As extensively reviewed by Ells
15
, the hydrides grow preferentially along the basal
plane of the HCP structure of -Zr. This has been exploitedby texture engineering the alloy
during manufacturing of tubular cladding in such a way that the normal to basal planes
of the zirconium grains (the < 0002 > direction) aligns along the radial axis of the tube
6,34
.
The end result is a series of hydrides aligned circumferentially (see g. 1.3), and therefore
hindering, to some extent, the propagation of radial cracks.
Growth direction of the hydrides will also be heavily inuenced by stress gradients.
There are currently two conicting models of how Delayed Hydride Cracking occurs:
the Diusion First Model (DFM) also known as the Dutton-Pulls model
2326
; and the
Precipitation First Model (PFM), strongly supported by Kim
2730
. Even though at the heart
of the debate is the diusion mechanism of hydrogen, both models infer that it diuses
up the stress gradient from compressive to tensile stress regions. This could be driven
by chemical potential dierence (DFM), which is inuenced by the stress eld, or by H
concentration (PFM). This means that at a stress concentration point (which could be a
4
(a) (b) (c)
Figure 1.2: The 3 hydride phases, courtesy of S. Lumley
22
. a) The metastable
phase has a fct Zr sublatice with
c
/a > 1 and only up to 0.5 of the interstitial sites
occupied. b) The hypo-stoichiometric phase has a fcc sublatice and it is mostly
found with a composition of ZrH
1.66
, therefore not all the tetragonal interstices
are occupied, up to the extreme case of a diamond-like structure. c) The phase
has an fct (
c
/a < 1) structure and occupancy of most of the tetragonal sites, up to
all in the case of ZrH
2
.
pre-existing crack tip, a surface defect, an impurity or another hydride) the hydrogen will
precipitate to form a hydride aligned in the direction of the crack. This step is followed by
rapid growth of the crack through the hard and brittle hydride. This leads to a situation
with the same conditions as the starting point, except eective crack size is increased by
the size of the hydride. DHC is a stable crack growth mechanism (constant crack growth
rate), until the critical stress concentration factor K
IC
is achieved for which the crack can
propagate through the ductile metal matrix, causing failure of the cladding.
DHC is a complex mechanism, with many inuencing factors, including minimum stress
concentration factor K
IH
, H solubility hysteresis and temperature dependence of crack
propagation. Owing to its complexity and the debate on the process involved, there is
no comprehensive review to date. For further details on the subject refer to the following
documents:
2333
1.3 zirconium Intermetallics
In the previous section, the eects of zirconium hydrides were discussed; however, one
should also consider the origin of hydrogen in the cladding. There are three distinct, but
nonetheless related, sources of H:
i The acidic environment in which the reactor operates. The coolant of water reactors is
eectively a slightly acidic solution as a result of a number of chemical additions. The
5
Figure 1.3: Typical Hydrides orientation in Zircalloy cladding. From Linga
Murty
6
.
most relevant are: boric acid, to control the neutronics of operation in a more uniform
fashion, as opposed to discrete control rods, which may distort the neutron ux in the
upper region of the core more than the bottom (or the other way round in the case
of BWRs); lithium hydroxide, mostly to counterbalance the decrease in pH caused by
the boric acid, but limited by the increase of corrosion rates of cladding material in
presence of Li
+
; and H
2
gas, added in the cold end of the primary circuit to scavenge
excess oxygen and help maintain the reducing environment.
ii Radiolysis of water molecules into aggressive free radicals. The products of radiolysis
may react with other water molecules or impurities (including organic molecules) in
the water stream, or may be subjected to further slicing from radiation. Examples of
such species are: OH, H, HO
2
, H
3
O
+
, O

2
, OH

, e

aq
, H
2
O
2
, H
2
, O
2
35
.
iii Oxidisation of the cladding surface. When, for example, a surface atomof Zr is oxidised
into ZrO
2
by two molecules of H
2
O, four atoms of H are released. These may dissolve
back into the water solution, recombine to formH
2
gas molecules, or be adsorbedby the
Zr metal and diuse into the bulk of the cladding. Equation 1.1 describe this behaviour
(from Lelievre
36
).
_
1 +
4y
x
_
Zr + 2H
2
O ZrO
2
4y
x
ZrH
x
+ 2
_
1 y
_
H

2
(1.1)
where y tends to be very small (i.e. most of the hydrogen recombines).
As the oxide grows inward, mostly mediated by V

O
diusion
37
, the hydrogen has an
oxide layer of increasing thickness to diuse through before reaching the zirconium metal.
Whilst it is well known that the hydrogen diusivity in metals is very high, diusion
behaviour in Zirconia is less well understood despite having been the subject of many
studies
38,39
. It is generally accepted that the measured and calculated diusivity of H in
ZrO
2
is relatively low and does not predict the rate of H uptake in zirconium alloys. One
model currently the most agreed upon reviews the role of the alloying elements in
6
H uptake. As shown in table 1.1, all the common alloying agents have higher electro-
negativity compared to zirconium (1.33 on the Pauling scale). This implies that if some
intermetallic phases formwithinthe zirconiummatrix (inthe formof SecondPhase Particles
or SPP) they will be less reactive to oxygen, and therefore oxidise at a lower rate, so that as
the oxide layer grows around them they retain a partially metallic form and act as a bridge
to H diusion
36,40
. Fig. 1.4 shows a schematic of this process (from Hatano
40
). From the
diagram one can also observe, according to this model, that larger SPPs exacerbate this
eect and lead to the mechanical failure of the cladding. This was subsequently conrmed
experimentally
11
. Furthermore, oxidation temperature will inuence the stability of the
intermetallic precipitates. At lower temperatures the SPPs will retain their metallic form
for longer, allowing increased absorption of H.
Various intermetallic phases have been identied in zirconium alloys, particularly for
Zircaloys as it contains a greater variety of alloying agents (see table 3.2 in the Results
section for a comprehensive list. However, it is worth pointing out that the presence and
composition of most intermetallic phases has been and still is subject to a great deal of
dispute. In the literature there are reports of conicting results regarding solid solubility of
Fe and Cr alloying additions in -Zr. Some papers report that all of the Fe and Cr content is
segregated in SPPs
41
, whilst others have shown that a certain degree of solubility in the Zr
matrix is retained
42,43
, and even that a supersaturated solution is obtained when quenched
from the phase
43
. Furthermore Fe-rich intermetallic precipitates have been observed
in some cases
42,44
, but not others
41,45
. Similar controversy surrounds Sn solubility. It is
well known that Sn has very large solid solubility in -Zr it is indeed an stabiliser
but there have been numerous reports of Sn-based SPPs with a very wide range of
compositions
42,46,47
. Note that Yang
45
has shown that redeposition of Sn onto precipitates
and grain boundaries during TEMsample preparation is a possible cause of misinterpreted
Sn segregation.
The issue of intermetallic precipitates in zirconium is obviously a complex one, and it is
made even more intricate by the dierent morphologies that each SPP may assume, and
by their subsequent relation and inuential growth on one-another
41
. For this reason
it is important to extend our knowledge of zirconium intermetallics, particularly their
interaction with H. Indeed hydrogen uptake might not be the only factor that is aected
by SPPs: hydride nucleation might be inuenced by the presence and morphology of the
intermetallics, as observed by Vitikainen
48
.
The necessity for further work in this area is evident. Clarication on the role of the
dierent alloying elements with regards to H uptake could lead to the development of
cladding alloys with greater reliability, and potentially also longer life expectancy, which
would accommodate fuel with higher burn-up. These are the foundations for the current
project.
7
Figure 1.4: Proposed mechanism for H diusion through the Zircaloy oxide
layer. (A) represents a zircaloy specimen that has been quenched from the
phase and annealed to produce uniformly distributed small intermetallic
particles, whilst (B) is slow cooled from the phase to produce fewer large
particles of similar composition
8
2
Methodology
In the past half-century the steady increase in computational power coupled with Moores
predictions
49
for the future of computation, which still hold true nowadays, have
encouraged the development of many algorithms and codes that investigate the behaviour
of materials at an atomic scale. Atomistic modelling is a eld in ever greater growth, that
is now also being largely adopted by industry. Its growing popularity is not only due to
the increase in the accessibility of computational power, but also because of its inherent
cost eectiveness (compared to large scale experiments), time saving nature (accelerated
modelling of processes), as well as the excellent rate of success.
Atomistic modelling is a eld that has been in continuous change since its birth, always
adopting newer and more accurate techniques and rejecting those that have become
obsolete. These transitions have always been driven by the availability of greater
computational power, and the present situation is no dierent. We are currently observing
the shift fromclassical potential models to quantummechanical ones. Because we are but at
the beginning of this process, and each model has its limitations, the two methods need to
coexist and supplement each-other in order to achieve a better understanding of the greater
picture. The current work has been carried out using quantum mechanical simulations,
as it is more suitable for the study of zirconium and zirconium intermetallic systems as
pointed out in the comprehensive reviews by Domain
5052
.
The main dierence between Quantum Mechanical (QM) simulations and empirical pair
potential simulations is that the latter represents ions as a whole entity (commonly as a hard
9
sphere shell and a spring), whilst the former simulates the behaviour of the electron cloud
that form the atom and denes its interactions with its surroundings. Quantum mechanics
is universally considered as the model that provides the most complete description of
nature, and it is the closest we have yet attained in answering the ultimate question of
Life the Universe and Everything. But in order to do so it inevitably carries with it a large
load of detail, and it is this complexity of information that prohibits the use of QM in the
representation of large systems. In fact the state of the art in QM simulation is limited to
complex molecules, or crystals of the order of 100 atoms (see convergence section for details
on the size limits of the current work).
2.1 Quantum Mechanics
At the root of quantum mechanics is the non-relativistic many body Schr odinger equation:

H(R
I
, r
i
) = E(R
I
, r
i
) (2.1)
Where is the wavefunction of the system as a function of ion positions (R
I
) and electron
positions (r
i
), E is the total energy of the system and

H is the Hamiltonian operator, which
contains the kinetic (T) and potential (V) energy contributions of all the ions and electrons
in the system:

H =

T
N
+

T
e
+

V
Ne
+

V
NN
+

V
ee
(2.2)
where

T
N
and

T
e
represent the kinetic energy components of the electrons and nuclei
respectively, and the remaining terms take into account the potential that arises due to
the ion-electron
_

V
Ne
_
, ion-ion
_

V
NN
_
and electron-electron
_

V
ee
_
interactions. These can be
mathematically expressed as follows (in the same order as in equation 2.2):

H
T
=

2
2M
I

2
R
I

2
2m
i

2
r
i
(2.3a)

H
V
=

I,i
Z
I
e
2
|R
I
r
i
|
+

I,J
Z
I
Z
J
e
2
2

R
I
R
J

i, j
e
2
2

r
i
r
j

(2.3b)
where R
I
, M
I
and r
i
and m
i
are the positions and masses of the I
th
ion and i
th
electron
respectively, is the Planks constant divided by 2 (also known as the Dirac constant), Z
I
and Z
J
are the atomic numbers of the I
th
and J
th
ion and e is one electric charge unit.
By minimising the LHS of equation 2.1 with respect to R and r it is possible to nd the
ground state of a system. Note, however, how the number of components within the
Hamiltonian operator may escalate quite rapidly for a relatively large system (10 for H
2
,
127 for methane, 279 for ethane), therefore making the computation of such systems quite
arduous.
10
2.2 Born-Oppenheimer Approximation
Due to the complexity of the full Hamiltonian operator, it is necessary to adopt a series
of approximations in order to compute the ground state of large systems. The rst of
such operations is the Born-Oppenheimer approximation: because the dierence in mass
betweenelectron and ions is very large (
m
e
/M
I
1), but the momenta are of the same order of
magnitude (
p
e
/p
I
1), the ions are eectively immobile with respect to the electrons and the
electronic response to ionic motion is eectively instantaneous. Therefore the wavefunction
can be separated into two independent component, an electronic one and an ionic one, to
be solved in distinct steps.
(r, R) =
e
(r; R) (R) (2.4)
where
e
is a function of r only, with the R dependence parameterized.
In the rst stepthe ions are consideredstationary (or clamped) which allows to the reduce
the Hamiltonian operator by ignoring the kinetic energy of ions

T
N
(being smaller than

T
e
by a factor of
M
I
/m
e
) and the ion-ion interaction potential

V
NN
(which is now constant and,
therefore, only provides an external potential that is felt by all electrons). This yields what is
known as the clamped nuclei approximation:

H
e
(r; R) =

T
e
(r) +

V
Ne
(r; R) +

V
ee
(r)
=

2
2m
i

2
r
i

I,i
Z
I
e
2
|R
I
r
i
|
+

i, j
e
2
2

r
i
r
j

(2.5)
and the respective Schr odinger equation to be solved becomes:

H
e
(r; R) e (r; R) = E
e
e (r; R) (2.6)
In the second stage, the nuclear kinetic energy term

T
N
, and the ion-ion interactions

V
NN
are reintroduced together with the electronic contribution to the total energy E
e
to solve the
nuclear part of the Schr odinger equation:

H
N
(R) (R) =E
tot
(R)
_

T
N
+

V
NN
(R) + E
e
_
(R) =E
tot
(R) (2.7)
2.3 Hartree-Fock Method
From the section above it is clear that the troublesome stage in the minimisation process is
to solve for the electronic contribution to the systems total energy. The starting point for
such calculation is to consider the electrons as non-interacting fermions. This is obviously
11
a very bold assumption, and the consequences of which are discussed below. For now lets
consider the case of a system with non interacting electrons, the electronic wavefunction
can be re-written as a Hartree product (after Douglas Hartree
?
):

HP
({r
i
}) =
n

i=1
(r
i
) (2.8)
where are the normalised wavefunctions of the individual electrons. The above equation
only considers the spacial variation of the electrons, neglecting the 4
th
degree of freedom:
the spin coordinate . It is easy to add the spin dependance by dening x = {r, } (usually
just as x = r x = r) so that equation 2.8 becomes:

HP
({x
i
}) =
n

i=1
(x
i
) (2.9)
Once the Hartree wavefunction is produced, it is possible to calculate the expectation value
of the electronic contribution to the energy of the system, using the usual probabilistic
interpretation of quantum mechanics. Exploiting the variational theorem, the expectation
value of the Hamiltonian, generated with any trial wavefunction, must be greater or equal
then the ground state of the system. Mathematically:
E
HP
=
_

HP

H
e

HP
_
E
e
(2.10)
In other words we are seeking for the minimum value of E
HP
. This means that we can start
with a trial set of wavefunctions
i
, which will output a corresponding set of
j
and then
iterate the process with the output as the new input until the two quantities are the self
consistent.
Even though the Hartree product is a very convenient form of the wavefunction, it has
one shortcoming : it does not generally satisfy the antisymmetry principle, for which the
wavefunction [must be] antisymmetric with respect to interchange of the coordinates
(including spin) of a pair of electrons
53
, or mathematically:

e
(x
1
, x
2
, . . . , x
n1
, x
n
) =
e
(x
n
, x
n1
, . . . , x
2
, x
1
) (2.11)
This is a very important principle in quantum mechanics (in fact the Pauli exclusion
principle is a corollary of it) andit is crucial to overcome this limitation. This canbe achieved
if the anti-symmentry condition is imposed by expressing the Hartree wavefunction as a
12
Slater determinants (after John C. Slater
54
):

HF
({r
i
}) =
1

N!

1
(x
1
)
1
(x
2
)
1
(x
N
)

2
(x
1
)
2
(x
2
)
2
(x
N
)
.
.
.
.
.
.
.
.
.
.
.
.

N
(x
1
)
N
(x
2
)
N
(x
N
)

(2.12)
which may be compressed using bra-ket (or Dirac) notation into

j

k
_
, since it is
always possible to construct a determinant from the vector components.
The new equation for the wavefunction is known as Hartree-Fock (HF), and the same
variational argument as before can be used to solve for the energy of the system, which
yields a similar solution but with an extra term: the exchange term. Essentially this means
that if the positions of two electrons are swapped, the sign of the determinant is changed
leaving

unaltered.
As mentioned above, the underlying assumption upon which the HF method was found
is that electrons are non-interacting entities in the system. This is very convenient
mathematically because it transforms the many-electron problem into a one electron
in a mean led problem but even with a good eld of choice it represents only very
poorly a many electron system. The exchange functional has introduced a relation between
parallel-spin electrons, but a more generic relationship between electrons was needed to
describe the Coulomb correlation eect.
2.4 Kohn-Sham and Density Functional Theory (DFT)
In 1964 Hohenberg and Kohn
55
suggested that the main problem of the HF approach is the
choice of fundamental variable: they believed that the many-electron wavefunction was
far too complex to be suitably approximated with only a few terms. The new starting point
proposed was the electron density, and for the validation of a new Hamiltonian equation
in terms of this variable they proved
1. by reductio ad absurdum that the external potential, generated by the stationary ions
(BO approximation), is a unique functional of electron density V
ext
= V
ext
[n (r)]; and
because the electron wavefunction depends on the external potential, it too must be
a functional of electron density = [n (r)];
2. that the groundstate energy, and all the following properties, can be found
variationally as a function of density.
13
From the above principles, Kohn and Sham, formulated a new Hamiltonian equation that
resembles Schr odinger equation but with ctitious non-interacting particles that behave
like (i.e. have the same density as) real interacting electrons:
E[n (r)] =
_

H
KS

_
(2.13)
Where the Hamiltonian operator is composed of the electronic kinetic energy term

T
e
, the
classic (Hartree) electron interaction potential

V
e
both of which can be expresseddirectly
as functions of electron density and the external potential

V
e f f
[n (r)]. By expanding
equation 2.13, the following is equations are obtained:
E[n (r)] =
_

T
e
+

V
e

_
+
_
V (r) n (r) dr
= F [n (r)] +
_
V (r) n (r) dr (2.14)
Fromthis point onwards, DFTproceeds in a similar fashion to the Hartree-Fock approach to
generate the self-consistent method for the minimisation of the energys expectation value.
The single-particle equation becomes:
_

2
2m
e

2
r
+ V
e f f
_
r, n (r)
_
_

i
(r) =
i

i
(r) (2.15)
where V
e f f
is the eective potential that acts upon the ctitious electron-like particles.
V
e f f
= V
e
+ V
XC
+ V
ext
(2.16)
and V
XC
is the exchange-correlation function. This term is the only one that cannot be
derived exactly ab-initio so further approximations must be made.
2.5 Approximations
The most common methods to estimate the XC function are the Linear Density Approxi-
mation (LDA) and the Generalised Gradient Approximation (GGA). The former simply
calculates the exchange-correlation potential from a uniform electron gas distribution.
Such a simple approximation works remarkably well in many systems, especially weakly
correlated ones, however, previous work from Domain
50
shows that for the zirconium
system the GGA works best. This takes into account the rst derivative of the electron
density variation in the system as a weighting factor for the exchange correlation eect.
Another very powerful approximation that allows DFT calculations of large systems and
in particular systems with heavy atoms is the pseudo-potential method, rst developed
14
by Phillips
56
and later adapted by Heine and Cohen
57
. When considering the chemical
and physical properties of materials, the interest is usually limited to the valence electrons
of the atoms, since the core electrons are so tightly bound by the nucleus that are eectively
invisible to the other atoms. Even though the contribution of the core electrons is important
the atom would behave dierently if those werent there it is only the eect that they
have upon the valence electrons that is relevant.
The pseudo-potential method allows to compute such approximation by curve-tting a
softer potential to the real potential of the electron to match only the outer part of the
wave function (which, as mentioned in the previous section, is a unique functional of the
external potential, which in turn depends on the ionic potential). See g. 2.1 for a graphical
representation.
Figure 2.1: The pseudopotential (red) is made to t the real potential (blue)
in such a way that the respective wavefunction is matched from innity to a
critical radius r
c
. Modied by A. Patel
58
, originally from Payne
59
.
Lastly, As we are interested in the bulk properties of materials, we employ periodic
boundary conditions. This approach consists in creating a unit-cell of the crystal of interest,
and then repeating this cell innitely in each direction. This procedure is possible thanks
to Blochs theorem, which allows to rewrite any wavefunction
i
(r) as the product of a
cell-periodic components f
i
(r) and a wavelike component e
ikr
:

i
(r) = f
i
(r) e
ikr
(2.17)
As a consequence the problem of solving innite wavefunctions is reduced to the problem
of solving a nite number of wavefunctions at an innite number of points in k-space.
However, because the electronic wavefunction at two k-points very close together are
15
almost identical, it is therefore possible to represent the wavefunction over a region in
k-space by that at a single k-point
59
. Therefore only a nite number of k-points is required
to calculate the ground state properties of an innitely large bulk of periodic crystal.
The great advantage of this approach is also its main limitation: it will periodically repeat
each and every point within the box boundaries. This implies that it approximates a perfect
crystal to near perfection, but for point defect calculations it will eectively produce a
defect concentration. To minimise the eect of defect-defect interaction one should use
the largest box possible, but this necessarily means having more atoms, therefore greater
computational power is required see g. 2.2.
0 20 40 60 80 100 120 140 160
0
2
4
6
8
10
12
14
Number of atoms in SC
D
e
f
e
c
t

i
n
t
e
r
a
c
t
i
o
n

d
i
s
t
a
n
c
e

(

)
Figure 2.2: The blue points are simulations of -Zr withaninterstitial hydrogen.
The size of the cell of Zr was increased from a single unit-cell (2 atoms) to a
553 supercell. The red line is not a best-t line, but the predicted behaviour
of y = ax
3
, where a is a geometrical factor. As the size of the supercell increases,
and therefore the linear defect-defect distance increases, the number of atoms
to compute increase according to a cube law. The reason why the points
are slightly o the predicted trend is mostly due to the anisotropy of the -
zirconium unit-cell
16
2.6 Defect Energy
In the previous sections it was explained how DFT allows calculation of the ground state
energy of a system. This section will briey discuss how the ground state energy is used to
determine the energy of point defects.
Consider the following incorporation reaction for a box containing 64 atoms of zirconium
(a 4 4 2 supercell):
1
2
H
2
_
g
_
+ Zr
64
(s) = Zr
64
H(s) (2.18)
The standard enthalpy of reaction is given by:
H

r
= H

f
(Zr
64
H) H

f
(Zr
64
)
1
2
H

f
(H) (2.19)
where H

f
is the standard enthalpy of formation. By denition the enthalpy of solution
E
sol
is the enthalpy change associated with the dissolution of a substance [H] in a solvent
[Zr
64
] at constant pressure, resulting in innite dilution
5
. This means that if the box
of -Zr were innitely large the above enthalpy of reaction would equate the enthalpy
of solution. And, as shown in the convergence section, it is reasonable to state that the
hydrogen interstitials are dilute non-interacting defects. In practical terms, this means that
the enthalpy of solution of a hydrogen interstitial may be calculated ab-initio by computing
the ground state energy for three systems, as shown in g. 2.3.
H
H
+
H
+
H
Figure 2.3: graphical representationof the incorporation reaction of a hydrogen
atom into a zirconium (left) or zirconium intermetallic (right) box as a series of
3 distinct simulations in each case.
The current work reports the solution energy as a comparison tool; what is of crucial
interest is the dierence in the enthalpy of solution between the pure zirconium and each
of the zirconium intermetallics. When subtracting one term to the other, the hydrogens
enthalpy of formation term cancels out; taking for example a box of 96 atoms of Zr
2
Fe as
the intermetallic phase:
E
H
sol
= E
H
sol
(IM) E
H
sol
(Zr) (2.20a)
= H

f
(Zr
64
Fe
32
H) H

f
(Zr
64
Fe
32
)
1
2
H

f
(H)

_
H

f
(Zr
64
H) H

f
(Zr
64
)
1
2
H

f
(H)
_ (2.20b)
= H

f
(Zr
64
Fe
32
H) H

f
(Zr
64
H) H

f
(Zr
64
Fe
32
) + H

f
(Zr
64
) (2.20c)
17
or, schematically:
+ +

H H
Zr Zr IM IM
Figure 2.4: Graphical representation of the series of simulations required to
calculate the dierence in solution energy E
sol
between two dierent solvents
(Zr and IM).
2.7 Nudged Elastic Band (NEB)
In the latter part of the project, the interest of the investigation moved towards the kinetics
of hydrogen diusion in zirconium. A useful parameter to quantify the diusivity of
hydrogen in zirconium is the activation energy for diusion, which corresponds to the
largest energy barrier between two stable minima. To extract such parameter, the Nudged
Elastic Band (NEB) method was adopted. This technique consists of a number of parallel
replicas (images) of the same system, with the defect in a dierent position between two
known minima. Each image is minimised independently, but during the ionic step the only
motion allowed is that along the directions orthonormal to the hyper-tangent, which is
calculated as the normal vector between two neighbouring images
60
. As a result, when the
simulation is completed, the chain of images form what is known as the Minimum Energy
Path (MEP), see g. 2.5.
Figure 2.5: An energy map around two local minima (red points). The
black dashed line represents the initial guess for the position of the images
(equidistant linear interpolation), and the blue solid line is the MEP. Modied
from QuantumWise
61
18
2.8 One last word about simulation vs experimental
Even though computer modelling of materials may be considered more predictive then
experimental works, it must be stressed that the results of modelling can only be validated
by direct (experimental) observations. For this reason a constant collaboration between
experimentalists and modellers should be sought.
2.9 general references for the methodology
...
19
3
Results
3.1 Convalidation of results
The rst part of the project involved validating the results of the project by proving the
convergence of the variable that have a signicant impact on the accuracy. These were
identied (in order of importance) as
1. the nature of the potential used,
2. the supercell size,
3. the cuto potential,
4. the k-space sampling density,
5. and the -smearing of the partial occupancies.
In the following chapter a brief outline of the convergence of each parameter is presented,
followed by a summary of the parameters used in the calculations
3.1.1 Choice of functional
Previous work from C. Domain
50
has showed that GGA potentials are more suitable than
LDA ones to investigate bulk properties of the Zirconium system. Furthermore, because
20
Fe Nb Sn Y (a,c) Zr (a,c)
Experimental
6265
() 2.865 3.293 6.4892 3.6474 5.7306 3.23 5.145
US-PP
LDA -11.50 -6.80 -3.32 -13.24 -25.26 8.50 -11.70
GGA 2.80 -0.10 14.38 -2.04 -11.26 -1.80 -2.10
PAW
LDA -13.10 -3.00 -1.12 -10.94 -20.06 -6.80 -6.80
PW91 -3.80 2.90 16.08 0.06 -6.56 0.40 1.50
PBE -3.30 3.30 16.58 1.06 -5.36 0.30 2.80
Table 3.1: Lattice constant calculation compared to experimental values. The
values for the 3
rd
row onwards are the percentage deviations from the reference
value (1
st
row). Modied from, and all calculations carried out by Lumley
22
.
Parameters used in calculations unknown.
we are interested in the bulk properties, it is intuitive to use Projector Augmented Wave
potentials rather than localised ones. This reduced the choice to only two options, both
PAW-GGA: Perdew-Wang 1991 (PW91) and the Perdew-Burke-Ernzerhof 2002 (PBE).
However, as mentioned in the 1, the current project is closely related to the work carried
out by S. Lumley in his PhD, and it was therefore important to have results that were
comparable.
In his work, Lumley produced a comparison of predicted versus experimental lattice
parameters for dierent functionals, see Table 3.1. From this comparison it was evinced
that the PW91 was the most appropriate potential.
3.1.2 Planewave cuto energy
When dealing with a problem of innite points, in practical terms one must resort to a
large but nite number of points to be able to solve it. This is the case with the basis set for
electrons in DFT (or any other QM computation method).
Even at a single k-point the solution to the Kohn-Sham equaiton is given by an innite
summation. Fortunately not all terms have the same weight, in fact each term represents
an energy level with a kinetic energy associated with it, and the higher the level the larger
the kinetic energy contribution. It is clear that the terms with the lowest kinetic energy have
the greatest inuence in the ground state of the system, it is therefore possible to dene an
arbitrary cut-o value for which any higher order solution will be neglected.
A suitable cut-o potential is one that minimises computational power associated with
the number of bands to compute whilst also providing a suciently large number of
empty levels to avoid artefacts in the results.
21
The graph in g. 3.1 shows the convergence for the planewave cut-o potential, with the
systems total energy. It is clear that if a precision of 10
3
eV/atomis sought than a potential
300 400 500 600 700
-6
-5
-4
-3
-2
-1
0
-Zr
-Zr
-ZrCr2
Zr3Fe
Zr2Fe
ZrFe2
Zr3Sn
Zr5Sn3
Cut-off potential (eV)
l
o
g
(
E
0
/
a
t
o
m
)
Figure 3.1: The abscissae axis represents the cut-o value (in eV) and the
ordinate axis shows the dierence in energy per atom between each point and
the previous one, logarithmically. All the calculations were carried using unit
cells with a very large (and constant) BZ sampling.
in the region of 450 eV should be used. However, because other parameters (e.g. the
supercell size) would be limited to a precision below 10
2
if such cut-o was adopted, it
was necessary to trade-o one order of magnitude of uncertainty here. It was therefore
chosen to use a potential cut-o of 350 eV, which produced a gain in computational power
of roughly 50%. Further analyses showed that this cut-o value limited the precision only
to 5 10
3
eV/atom, therefore still below the uncertainty obtained from defect-defect
interactions (see section 3.1.4). The self consistent loops were minimised to 10
4
eV, which
is a relatively inaccurate value, andthis explains the large scatter at lower values. Following
this, further results were calculated using more accurate energy stopping criteria for the
electronic relaxation. It is worth noticing that the speed of convergence is very element-
specic: comparing the curves for Zr
3
Fe, Zr
2
Fe and ZrFe
2
one can observe a simple shift
upwards (slower conversion) with increasing Fe content. Similarly the Zr-Sn intermetallics
have similar curves, and so do the pure Zr phases. It is perhaps not very surprising that
Fe and Cr are the most problematic species, given that they have the largest number of
electrons in the outer shell.
22
3.1.3 Brillouin zone sampling
As mentioned in the methodology section (), it is necessary to approximate the (smooth)
variation of the wavefunctions in k-space as a series of discrete solutions around distinct
points. The approximation works best for innitely close points, and degrades as the
distance between sampling points increases. However, denser sampling requires greater
computational power, making it prohibitive to use very large number of k-points.
The graph in g. 3.2 illustrates the increase in precision as the distance between k-points
decreases. It is clear that a precision of 10
3
eV/atom requires a sampling density of roughly
0 20 40 60
-5
-4
-3
-2
-1
-Zr
-Zr
-ZrCr2
-ZrCr2
Zr3Fe
Zr2Fe
ZrFe2
Zr3Sn
Zr5Sn3
linnear K-point density ()
l
o
g
(

E
0
/
a
t
o
m
)
Figure 3.2: The ordinate-axis represents the dierence in groundstate energy
per atom between each point an the one that precedes it (i.e. with fewer k-
points), logarithmically. Owing to the large scatter of the plot, it was decided
not to represent the individual convergence curve of each series, but rather the
collective behaviour of all phases. The shade was drown as a guide to the eye
only. Data obtained with unit cells of each phase, xed cut-o potential of 450
eV, and altering the Monkhorst-Pack grid by increasing the least dense vector
rst. -centering used for hexagonal, rhombohedral and triclinic cells.
60 k-points , equivalent to k-point distance of 0.01

6
1
. However, a precision inferior by
one order of magnitude will be easily achieved around 25 k-points (0.04
1
in reciprocal
spacing). As a consequence of this, and considering the grids that would have been optimal
23
for the relevant supercells, a k-point spacing of 0.03
1
was opted for (linear density
equivalent is 33 ). The decision was done in light of the fact some Monkhorst-Pack
grids are more ecient than other. For instance, for cubic, tetragonal and orthorhombic
systems, non--centredeven number grids are more ecient than oddnumber or -centred
ones. This is because whenthe Brillouin Zone is reducedtoits irreducible constituent, many
of the k-points (upto
1
2) will become redundant by symmetry, thus reducing computational
power needed for the same degree of accuracy. However, if the k-points are not positioned
symmetrically to the symmetry axis (as is the case for -centred grids in a square system),
then the total number of k-points in the IBZ is only a little less than that of the full BZ. As
a direct consequence, decreasing the number of k-points (and therefore accuracy) does not
necessarily mean a decrease in computational power: a 3 3 3 grid will contain more
points in the IBZ than a 4 4 4 grid for a square system.
To summarise the grid type used:
for

, ,

= 90

evennumber MPgrids were adoptedupto8points per reciprocal vector,

then odd number ones were employed (starting from 11).
for

, ,

< 60

, ,

> 120

-centred grids were used, were possible even

numbers below 8 and odd numbers above 11.
3.1.4 Defect concentration limit
As mentioned in the Methodology section 3.1.3, it is very computationally expensive to
model larger supercells, however because we intend to model isolated point defect energy,
it is vital to minimise the defect-defect interaction. The largest supercells that possible
to compute contained 150 atoms, equivalent t a H concentration of 73.6 wt. ppm, which
is just above the solid solubility of H in -Zr at operating temperature. It is clear then
that supercell size should be kept as large as possible. In support of this, the variation in
groundstate energy with supercell size is of the order of 0.02eV for Zirconium (see table 3.4
and 3.5), which is much larger than the precision limit imposed by the other parameters
described in this chapter.
3.1.5 -smearing of partial occupancies
Partial occupancies are commonly used in DFT calculation since they reduce the number
of k-points required for the calculation of an accurate calculation band-structure energy
60
.
The smearing method used was the rst order Methfessel-Paxton, which produces similar
results to simple Gaussian smoothing but with much lower entropy terms. Ideally,
the ctitious temperature smearing parameter should be as large as possible whilst
maintaining the entropy value (which can be considered as the magnitude of uncertainty)
24
negligible. From the above analyses it is clear that a source of error can be dened as
negligible if it is less than 10
3
eV/atom. Fig. 3.3 shows the variation entropy component
of the total energy as a function of . It is clear from the above graph that if a precision
0 0.05 0.1 0.15
-6
-5
-4
-3
-2
-Zr
-Zr
-ZrCr2
Zr3Fe
Zr2Fe
ZrFe2
Zr3Sn
Zr5Sn3
(K)
l
o
g
(

S
)
Figure 3.3: The ordinate-axis represents the dierence in entropy per atom
between each point an the one that precedes it, logarithmically. The shade was
drown as a guide to the eye only.
of 10
3
is desired, than should be less than 0.04 K. However it was later discovered that
the increase in computational power between 0.01 and 0.04 K was marginal, and therefore
decided to retain the original value used by S. Lumley in his work.
25
3.2 Summary of intermetallic phases
Various intermetallic phases have been identied from the literature as relevant ones for
the current work, and they are summarised in Table 3.2. When examining the intermetallic
phases it is useful to refer to the relevant phase diagram, which are attached in the
Appendix ??.
Table 3.2: Summary of the main intermetallic phases found in Zirconium alloys
Alloy Phase Composition Occurrence and Notes
Zr (Fe,Cr)
2
HCP: ...
_
-ZrCr
2
_
ZrCr
1.1
Fe
0.9
41,46
Zr (Fe
0.75
Cr
0.25
)
2
66
Zr (Fe
0.6
Cr
0.4
)
2
45
Zr (Fe
0.4
Cr
0.6
)
2
45
Zr
x
Cr
2
Fe
5
46,67
Zr
0.36
Cr
.43
Fe
.21
68
Zr
0.32
Cr
.24
Fe
.44
68
Very common in Zy-2 and Zy-4
(often with dierent composi-
tions)
FCC: ...
(-ZrCr
2
)
46,48,69
Zr (Fe
0.1
Cr
0.9
)
2
44
Foundonly within grains, whilst
inthe same experiments the HCP
structure was found between
grains. Especially abbundant in
Scanuk alloys.
Zr
2
(Fe,Ni) BCT: ...
(Zr
2
Ni)
Zr
2
Ni
0.4
Fe
0.6
41,46,48
Zr
2
(Fe
0.5
Ni
0.5
)
45
Zr
2
(Fe
0.8
Ni
0.2
)
45
Very common in Zy-2
Zr
4
(Fe,Cr) FCC: ...
(Zr
4
Fe)
Zr
4
Fe
0.7
Cr
0.3
45
Zr
4
Fe
44
Metastable form found to dec-
orate the boundaries of the
lamellae formed in -quenched
Zircaloy
Zr-Sn ZrSn, Zr
5
Sn
3
, Zr
3
Sn,
Zr
5
Sn
4
, Zr
4
Sn, BCTSn
42,44,46,47,67
Oftenwithlarge Fe content. May
be an artefact of TEM sample
preparation
45
.
Zr (Fe
x
M
1x
)
2
HCP (x < 0.8),
FCC (x > 0.9)
M=V,Cr,Mn
70
Zr-Nb -Nb Nb is a stabiliser for Zr.
Of the all intermetallic phases, the only ones that are universally accepted to be present in
the relevant alloys are Zr(Fe,Cr)
2
and Zr
2
(Fe,Ni), and their composition has been identied
to a condent degree of certainty. Because Nickel has been discarded as an alloying agent
in Zircalloy-4 onwards (see table 1.1), the current study has focused on the intermetallic
phases not containing Ni.
26
3.3 Pure Zirconium
Zirconium metal has two stable phases, and , represented in g. 4.5 and described in
Table 3.3.
(a) -Zr (b) -Zr
Figure 3.4: Graphical representation of the unit cells of the two
zirconium phases with relevant intersticial sites. (a) The intersticials and
their crystallographic positions of -Zr are (Clockwise from the bottom
left): hexahedral
_
1
3
,
2
3
, 0
_
, tetrahedral
_
1
3
,
2
3
,
1
8
_
, non-basal trigonal
_
5
9
,
1
9
,
1
6
_
,
octahedral
_
2
3
,
1
3
,
1
4
_
and basal trigonal
_
2
3
,
1
3
, 0
_
. Notice that whilst the trigonal
site repeats every {0001} plane, the hexahedral repeats only every {0002} plane,
since a Zr atoms occupy the
_
1
3
,
2
3
,
1
2
_
site. (b) The intersticial sites for -
Zr are (from top to bottom): octahedral
_
1
2
, 0,
1
2
_
, tetrahedral
_
1
2
, 0,
1
4
_
and
trigonal
_
1
6
,
1
2
,
1
6
_
Table 3.3 also contains the calculated lattice parameters for both phases compared to
experimental work. It is clear that there is a decisive agreement between predicted and
experiment values, which condence for subsequent results.
The enthalpy of formation (E
sol
) and the defect volume (V) were calculated for each of the
defect sites described above; they are summarised in Table ??.
27
Table 3.3: Calculated lattice parameters compared to experimental values. The
uncertainty in the predicted lattice parameter derives from the variation of the
results with supercell size.
Phase Lattice type Space group Calculated Reference
(Int. table n.) a
0
c
0
a
0
c
0
-Zr HCP P6
3
/mmc (194) 3.233 0.003 5.162 0.011 3.23
62
5.145
62
3.2414
71
5.1696
71
3.231
72
5.142
72
-Zr BBC Im

3m (299) 3.570 0.001 3.5453

73
3.488
74
3.6162
75
Table 3.4: -Zr: Energy of solution (E
sol
) and Defect volumes (V) for H in all
the stable interstitial sites, for dierent supercell sizes.
Intersticial
Site
SuperCell
size
Atoms in SC
([H] wt. ppm)
r(H-H) in E
sol
V
Tet
2 2 1 8 (1380) 5.23 -0.427 3.36
2 2 2 16 (690) 6.48 -0.449 3.61
3 3 2 32 (345) 9.71 -0.463 3.20
4 4 2 64 (173) 10.34 -0.442 3.44
4 4 3 96 (115) 12.93 -0.474 2.13
5 5 3 150 (73.6) 15.47 -0.467 3.07
Oct
2 2 1 8 (1380) 5.18 -0.436 0.93
2 2 2 16 (690) 6.47 -0.465 1.13
3 3 2 32 (340) 9.71 -0.438 0.82
4 4 2 64 (173) 10.33 -0.404 1.05
4 4 3 96 (115) 12.92 -0.427 0.26
5 5 3 150 (73.6) 15.47 -0.432 0.76
Hex
4 4 2 64 (173) 10.31 -0.341 3.77
5 5 3 150 (73.6) 15.46 -0.356 3.40
Tri
4 4 2 64 (173) 10.34 0.018 4.49
5 5 3 150 (73.6) 15.48 0.074 5.34
nbT 4 4 2 64 (173) 10.34 -0.040 3.79
28
Table 3.5: -Zr: Energy of solution (E
sol
) and Defect volumes (V) for H in all
the stable interstitial sites, for dierent supercell sizes.
Intersticial
Site
SuperCell
size
Atoms in SC
([H] wt. ppm)
r(H-H) in E
sol
V
Tet
2 2 2 16 (690) 7.10 -0.644 7.3
3 3 3 54 (204) 10.59 -0.865 5.35
4 4 4 128 (86.3) 14.26 -0.650 3.16
Oct
2 2 2 16 (690) 6.85 -0.770 3.04
3 3 3 54 (204) 10.58 -0.699 9.25
4 4 4 128 (86.3) 14.25 -0.501 4.13
Tri
2 2 2 (204)
3 3 3 (86.3)
3
29
3.4 Zr-Fe intermetallics
(a) Zr
3
Fe (b) Zr
2
Fe
(c) ZrFe
2
Figure 3.5: Graphical representation of the unit cells of the three Zr-Fe
intermetallic phases.
30
In the Zr-Fe phase diagram (see ??), there are three stable intermetallic phases: Zr
3
Fe, Zr
2
Fe
and ZrFe
2
. Graphical representation of these are presented in g 3.6. Furthermore, there is
also a metastable form, with a stoichiometry of Zr
4
Fe, that has been documented in various
Zy-4 studies
44,45
, mostly withsome degree of Cr solutioninthe Fe site. However insucient
crystallographic data was found regarding the Zr
4
Fe structure in order to perform reliable
DFT simulations.
Values for the three phases that were investigated, with comparison to experimental data,
are summarised in Table 3.6.
Table 3.6: Summary of the Fe-Zr intermetallic phases with computed lattice
parameters and comparison with experimental data.
Phase Space group Calculated Experimental Reference
(Int. table n.) a
0
b
0
c
0
a
0
b
0
c
0
Zr
3
Fe Cmcm (63) 3.258 10.901 8.996 3.33 10.95 8.82 [Ardisson91]
76
3.324 10.974 8.821 [Yartys99]
77
3.32 11.00 8.82 [Stein02]
72
Zr
2
Fe I4/mcm (140) 6.268 5.720 6.41 5.56 [Ardisson91]
76
6.379 5.599 [Yartys99]
77
6.382 5.602 [Stein02]
72
ZrFe
2
Fd

3m (227) 7.011 7.016 [Stein02]

72
7.1 [Gorria04]
78
7.064 [Zotov08]
79
Defect energy calculations were performed using the following supercell parameters:
Zr
3
Fe: SC size= 3 1 2; n. atoms: 69; [H] = 127.4wt. ppm; r (H-H) = 9.78.
Zr
2
Fe: SC size= 2 2 2; n. atoms: 96; [H] = 132.2wt. ppm; r (H-H) = 11.44.
ZrFe
2
: SC size= 2 2 2; n. atoms: 48; [H] = 310.4wt. ppm; r (H-H) = 9.91.
ZrFe
2
: SC size= 3 3 3; n.atoms:162; [H] = 91.98wt. ppm; r (H-H) = 14.89.
where SC stands for SuperCell, and r (H-H) is the smallest defect interaction distance
between cell replicas. The resulting solution energies for hydrogen in the various stable
interstitial sites are presented in Tables 3.7. A negative E
sol
means that the site is more
stable than the lowest energy interstice in -Zr. Many more sites were investigated than the
ones presented in table 3.7, however they were found to decay into one of the more stable
sites.
31
Table 3.7: List of the stable interstitial sites for Zr
3
Fe, Zr
2
Fe and ZrFe
2
. E
sol
is
the dierence in solution energy of H with respect to a tetrahedral interstitial in
-Zr of similar supercell size.
Phase Wycko n. Type Coordinates E
sol
E
sol
Zr
3
Fe 4b Oct 0,
1
2
, 0 0.714 0.239
8f Oct 0, 0.035,
5
8
0.690 0.215
4c Tet 0, 0.035,
3
4
0.521 0.046
8f Tet 0, 0.324, 0.452 0.521 0.017
8f Tet 0, 0.279, 0.558 0.415 0.060
4c Oct 0, 0.614,
1
4
0.376 0.098
8f Tet 0, 0.117, 0.665 0.295 0.180
16h Tet 0.223, 0.302, 0.388 0.190 0.284
4c Hex 0, 0.242,
1
4
0.128 0.347
Zr
2
Fe 16l Tet
7
8
,
3
8
,
3
8
0.492 0.018
32m Tet 0.59, 0.27, 0.18 0.460 0.014
16j Tri
1
4
, 0,
1
4
0.449 0.025
4b Tet 0,
1
2
,
1
4
0.399 0.075
16k Tet 0.345, 0.452, 0 0.258 0.217
4c Lin 0, 0,
1
2
0.225 0.249
ZrFe
2
96g Tet 0.422, 0.422, 0.220 0.110 0.563
32e Tet 0.385, 0.385, 0.385 0.128 0.580
48f Tri 0.206,
1
2
,
1
2
0.202 0.654
8b Tet
1
2
,
1
2
,
1
2
0.236 0.689
96g Tri 0.112, 0.112, 0.319 0.237 0.691
32e Tri 0.792, 0.792, 0.792 0.378 0.831
16c Lin
3
8
,
1
8
,
3
8
1.327 1.780
32
3.5 Zr-Cr intermetallics
(a) -ZCr
2
(b) -ZCr
2
(c) -ZCr
2
Figure 3.6: Graphical representation of the unit cells of the three Zr-Fe
intermetallic phases.
The Zr-Cr binary system (see ??) has only one stoichiometry for intermetallic compounds
(1:2), however, within that stoichimetry there are three phases. They are all layered Laves
33
phase with dierent stacking sequences (c for cubic and h for hexaagonal): -ZrCr
2
has
the C15 Laves structure composed of three layers with a ccc stacking sequence; -ZrCr
2
has the C36 Laves structure composed of four layers with a chch stacking sequence; and
-ZrCr
2
has the C14 Laves structure composed of two layers with a hh stacking sequence
80
.
It must be stressed that even though the -ZrCr
2
phase hardly appears in the literature
around zirconium alloys, XRD studies have shown that the C36 structure is often present
in crystals of C14 as a concentration of staking faults in the layered sequence
80
.
Table 3.8: Summary of the Cr-Zr intermetallic phases with computed lattice
parameters and comparison with experimental data.

Values with an asterisk are taken from commercial alloys such as Zircaloy-2,
Zircaloy-4 and Scanuk.
Phase Space group Calculated Experimental Reference
(Laves structure) (Int. table n.) a
0
c
0
a
0
c
0
-ZrCr
2
(C15) Fd

3m (227) 7.118 7.204 [Soubeyroux95]

81
7.23 [Gonzalez05]
82
7.4

[Vitikainen78]
48
7.19

[Krasevec81]
69
7.21

[Yang87]
44
-ZrCr
2
(C36) P6
3
mmc (194) 5.06 16.29 5.11 16.56 ??
-ZrCr
2
(C14) P6
3
mmc (194) 5.09 8.04 5.113 8.309 [Soubeyroux95]
81
5.103 8.268 [Mestink Filho]
80
5.079

8.279

[Vander Sande]
67
5.01

8.22

[Krasevec81]
69
5.09

8.27

[Yang87]
44
Defect energy calculations were performed using the following supercell parameters:
-ZrCr
2
SC size = 2 2 2; n. atoms = 48; [H] = 322.6wt. ppm; r (H-H) = 10.07.
-ZrCr
2
SC size = 3 3 3; n. atoms = 162; [H] = 95.60wt. ppm; r (H-H) = .
-ZrCr
2
SC size = 2 2 1; n. atoms = 96; [H] = 161.3wt. ppm; r (H-H) = .
-ZrCr
2
SC size = 2 2 2; n. atoms = 96; [H] = 161.3wt. ppm; r (H-H) = 10.16.
where SC stands for SuperCell, and r (H-H) is the smallest defect interaction distance
between cell replicas. The resulting solution energies for hydrogen in the various stable
interstitial sites are presented in Tables 3.9. As for the Zr-Fe case, more sites were
investigated than the ones presented in this report but they were found to be unstable,
and, hence, decay into one of the sites described in Table 3.9.
34
Table 3.9: List of the stable interstitial sites for -ZrCr
2
, -ZrCr
2
and -ZrCr
2
.
E
sol
is the dierence in solution energy of H with respect to a tetrahedral
interstitial in -Zr of similar supercell size.
Phase Wycko n. Type Coordinates E
sol
E
sol
-ZrCr
2
96g Tet 0.422, 0.422, 0.220 0.241 0.211
32e Tet 0.385, 0.385, 0.385 0.204 0.248
48f Tri 0.206,
1
2
,
1
2
0.060 0.392
8b Tet
1
2
,
1
2
,
1
2
0.058 0.394
16c Lin
3
8
,
1
8
,
3
8
1.212 1.664
-ZrCr
2
6h Tet 0.426, 0.573,
1
4
0.277 0.191
4f Tet
1
3
,
2
3
,
7
16
0.234 0.234
12k Tet 0.533, 0.467, 0.409 0.227 0.241
24l Tet 0.384, 0.351, 0.454 0.185 0.283
6h Tet 0.245, 0.490,
1
4
0.172 0.296
12k Tet 0.161, 0.322, 0.128 0.157 0.310
4e Tet 0, 0,
1
8
0.128 0.595
2c Lin
1
3
,
2
3
,
1
4
1.262 1.730
35
3.6 Zr-Sn intermetallics
(a) Zr
3
Sn (b)
Zr
2
Fe
Figure 3.7: Graphical representation of the unit cells of the two stoichiometric
Zr-Sn intermetallic phases that were analysed in the current work.
The zirconium-tin binary system comprises of various phases of dubious nature. Much
controversy surrounds the composition and more importantly the structure of the
Zr-Sn intermetallic phases. The two most predominant intermetallics in the binary system
present partial occupancy or self-interstitial occupancy of lattice sites: the Zr
4
Sn phase
derives from the Zr
3
Sn structure with
1
4 of the Sn sites being occupied by a Zr atom; the
Zr
5
Sn
3
phase tends to have self-interstitial occupancy of Sn up to a stoichiometry of Zr
5
Sn
4
.
As a limitation of DFT, we are unable to compute such structures, however in the current
work the structures with no excess or defect Sn are investigated to provide a platform for
further analyses with more suitable tools (such as quasi-random techniques). These are
summarised in table 3.10. Another structure that has been reported in the literature is
the orthorhombic ZrSn
83
, however insucient crystallographic data is available to model
its structure reliably with DFT. Furthermore the tin-rich side of the binary system (and
therefore the ZrSn
2
intermetallic) has been disregarded since such it appears not to be
relevant for zirconium alloys.
Defect energy calculations were performed using the following supercell parameters:
Zr
3
Sn SC size = 2 2 2; n. atoms = 64; [H] = 160.5wt. ppm; r (H-H) = 11.27.
36
Table 3.10: Summary of the Sn-Zr intermetallic phases with computed lattice
parameters and comparison with experimental data.
,
Values accompanied by the asterisk and dagger symbols were taken from the
Zr
4
Sn and Zr
5
Sn
4
stoichiometry respectively.
Phase Space group Calculated Experimental Reference
(Int. table n.) a
0
c
0
a
0
c
0
ZrSn
3
Pm

3n (223) 5.633 5.63 [Roteutscher65]

84
5.625

[Kwon90]
85
5.631

[Luo70]
86
Zr
3
Sn
5
P6
3
mcm (193) 8.538 5.799 8.456 5.779 [Kwon90]
85
8.462 5.797 [Gunnar60]
83
8.7656

5.937

[Kwon90]
85
Zr
5
Sn
3
SC size = 2 2 2; n. atoms = 128; [H] = 77.56wt. ppm; r (H-H) = 11.58.
where SC stands for SuperCell, and r (H-H) is the defect interaction distance between cell
replicas. The resulting solution energies for hydrogen in the various stable interstitial sites
are presented in Tables 3.11. Once again, the sites that were found to be unstable are not
presented in the report.
37
Table 3.11: List of the stable interstitial sites for Zr
3
Sn and Zr
5
Sn
3
. E
sol
is the
dierence in solution energy of H with respect to a tetrahedral interstitial in
-Zr of similar supercell size.

This particular site is actually shifted o the trigonal plane towards the
octahedral site 2b, this is apparently more stable than the full octahedral site
(multiple simulations with dierent starting guesses were carried out to conrm
this).
Phase Wycko n. Type Coordinates E
sol
E
sol
Zr
3
Sn 6d Tet
1
4
,
1
2
, 0 0.662 0.220
8e Hex
1
4
,
1
4
,
1
4
0.376 0.065
48l Tri 0.273, 0.348, 0.424 0.100 0.342
6b Lin 0,
1
2
,
1
2
0.735 1.177
Zr
5
Sn
3
2a Tri 0, 0,
1
4
0.541 0.074
12k Tri

0.095, 0.553, 0 0.437 0.030

2b Oct 0, 0, 0 0.417 0.050
12i Hex 0.1955, 0.391, 0 0.018 0.449
12j Hex 0.210, 0.514,
1
4
0.051 0.518
24l Hex 0.461, 0.143, 0.402 0.464 0.931
12k Tri 0.430, 0, 0.430 0.909 1.376
12i Tri 0.454, 0.908, 0 1.035 1.502
6f Lin
1
2
, 0, 0 1.076 1.543
4c Lin
1
3
,
2
3
,
1
4
1.574 2.04
38
3.7 Zr-V intermetallics
(a) ZrV
2
Figure 3.8: Graphical representation of the unit cells of the only Zr-V
intermetallic compound.
The vanadium-zirconium system comprises of only one intermetallic compound: ZrV
2
.
This has the same crystal structure as ZrFe
2
and -ZrCr
2
described above. The calculated
and experimental values of the intermetallic are presented in Table 3.12.
Table 3.12: Details of the only Zr-V intermetallic phase with computed lattica
parameter and comparison with experimental values.
Phase Space group Calculated a
0
Experimental a
0
Reference
ZrV
2
Fd

3m (227) 7.324 7.443 [Lototsky05]

71
7.44 [Daumer88]
87
7.441 [Lawson78]
88
Defect energy calculations were performed using the following supercell parameters:
ZrV
2
SC size = 2 2 2; n. atoms = 48; [H] = 326.1wt. ppm; r (H-H) = 10.16.
ZrV
2
SC size = 3 3 3; n. atoms = 162; [H] = 96.65wt. ppm; r (H-H) = 15.27.
where SC stands for SuperCell, and r (H-H) is the defect interaction distance between cell
39
replicas. The resulting solution energies for hydrogen in the various stable interstitial sites
are presented in Tables 3.13.
Table 3.13: List of the stable interstitial sites for ZrV
2
. E
sol
is the dierence in
solution energy of H with respect to a tetrahedral interstitial in -Zr of similar
supercell size.
Phase Wycko n. Type Coordinates E
sol
E
sol
-ZrCr
2
96g Tet 0.422, 0.422, 0.220 0.717 0.249
32e Tet 0.385, 0.385, 0.385 0.644 0.176
96h Tri
1
8
, 0.066, 0.184 0.602 0.135
48f Tri 0.206,
1
2
,
1
2
0.575 0.107
8b Tet
1
2
,
1
2
,
1
2
0.360 0.107
32e Tri 0.792, 0.792, 0.792 0.307 0.161
48f Lin 0.375,
1
2
,
1
2
0.064 0.532
16c Lin
3
8
,
1
8
,
3
8
0.197 0.666
40
3.8 Nudged Elastic Band for -Zr
In aid of clarity, the interstitial sites, described in the 3.3 section, will identied as follows:
TET for the 4f tetrahedral site at
1
3
,
2
3
,
1
8
;
HEX for the ?? trigonal site at
1
3
,
2
3
, 0;
OCT for the 2d octahedral site at
2
3
,
1
3
,
1
4
;
TRI for the ?? trigonal site at
2
3
,
1
3
, 0;
nbT for the k? non-basal trigonal site at
5
9
,
1
9
,
1
6
.
The Nudged Elastic Band (NEB) method was used to calculate the migration enthalpy for
hydrogen in -Zr. the three resulting curves are shown in g. 3.9.
-1 -0.5 0 0.5 1
0
0.1
0.2
0.3
0.4
0
2
4
6
8
10
Distance ()

H

(
e
V
)

H

(
k
c
a
l
/
m
o
l
)
HEX
TET TET OCT
OCT
OCT
TRI
nbT
TET
Figure 3.9: NEB curves for the three possible migration mechanisms. The
curves have been centred at the saddle point and enthalpy change values (the
ordinate axis) were taken with respect to the lowest energy point on the curve.
Dashed lines indicate that the data was collected for only half the bell curve
exploiting the symmetry property of the migration process.
41
It is worthmentioningthat the validityof the symmetryoperationwas provedbycomparing
the above curves to ones obtained frommodelling the full migration process, which yielded
similar results, but with poorer resolution (same number of points over twice the spread).
The three migration processes occur in dierent directions and are shown in g. 3.9 in
such a way that the x-axis represents the distance from the saddle point, regardless of
the direction. As a consequence of the alignment, the same TET site is found at dierent
abscissae points. Furthermore, because each curve is normalised with respect to the most
stable site withinthe curve, the graphis not representative of the depthof the wells nor of the
dierence in stability between TET and OCT sites. However, this normalisation allows to
compare energy barrier height between dierent migration processes. The energy barriers
are presented in table 3.14.
Table 3.14: Energy barriers for diusion in the three cases described above,
with comparison to experimental data.
Migration process H (eV) H (cal/mol)
TET-HEX-TET 0.0997 2298.5
TET-nbT-OCT 0.4001 9228.6
OCT-TRI-OCT 0.4227 9748.3
Sawatzky1960 0.3634 0.0173 8380 400
89
Kearns1972 0.4618 0.0117 10650
90
42
4
Discussion
The literature contains various results (partially conicting) regarding the interstitial
occupancy of H in zirconium: for the phase, there is a general consensus that H is
more stable in the tetrahedral site rather than the octahedral one
17,19,20,50?
. For the phase
however, Zuzek
17
reports octahedral occupancy, while Setoyama
20
, Dupin
19
andDomain
50
assert a preference for the tetrahedral site. Even though the current work is in agreement
for the -Zr, it is a very small preference (of the order of 0.04eV) and it is realistic to believe
that H might occupy both sites. In fact, as discussed below, it is a requirement for diusion
to occur in -Zr. As for the phase, the dierence in solution energy becomes greater
( 0.1eV), but the scatter of points (with dierent cell sizes) is too large too conclude which
site is preferred: for smaller supercell sizes the tetrahedral site is more stable, but for the
supercell containing 128 atoms (which was problematic to converge), the trend reverses.
Note that the latter simulation was the least reliable.
The results for pure hydrogen showed that in the phase, the is no net preference between
the two sites.
Due to the large number of dierent intermetallic systems each with a dierent set of
stoichiometries, christallographic structures, relative radius sizes and electro-negativities
the comparison across the various systems is very arduous. Table 4.1 summarises the
solution energy of the most stable site only from each of the phases investigated. The rst
surprising feature of the results is that hydrogen exhibits a lesser preference for the ZrFe
2
and all ZrCr
2
phases compared to -Zr. As discussed in the introduction section 1.3, these
43
Table 4.1: The dierence in solution energy is taken with respect to the -Zr
TET site (with a similar defect concentration)
Phase E
sol
E
sol
V
-Zr -0.467 3.07
-Zr -0.644 -0.202 2.71
Zr
3
Fe -0.744 -0.239 0.61
Zr
2
Fe -0.492 -0.018 2.48
ZrFe
2
0.110 0.563 3.09
-ZrCr
2
-0.241 0.211 2.66
-ZrCr
2
-0.336 0.132 3.96
-ZrCr
2
-0.277 0.191 3.79
ZrSn
3
-0.662 -0.220 -0.18
Zr
5
Sn
3
-0.541 -0.074 1.87
ZrV
2
-0.717 -0.249 2.94
phases are thought to be the main SPP responsible for H uptake in Zircaloy-4. Because of
their higher electro-negativity, they retain a partially metallic form during the oxidization
of the zirconium alloy, providing a bridge for H diusion through the zirconia layer. The
current results provide the reason for which hydrogen would migrate out of the SPP into
the bulk Zr. Furthermore, it implies that it will most likely be retained in the Zr matrix,
rather than segregate into the Zr(Fe,Cr)
2
particles. These results are particularly relevant
for Zircaloy-4 where such SPP is very common.
4.1 Electro-negativity
Various attempts have been made to correlate the stability of H interstitials to the physical
and chemical properties of the various phases. The rst of which was the electro-negativity
of the alloying metal compared to zirconium. This seems like an intuitive comparison
tool, which should reveal whether the chemical anity is the predominant process in the
solubility of hydrogen in the metals. However there seem to be little correlation between
solution enthalpy and electro-negativity of the alloying metals, as one can observe from g.
4.1. The graph only shows the lowest energy sites for each phase, since it is reasonable to
assume that the H interstitial will chiey occupy the most stable congurations.
For those binary systems that have multiple intermetallic stoichiometries, namely the
Zr-Fe and the Zr-Sn system, it is possible to identify an uprising trend with increase in
electro-negativity. However, this apparent trend is heavily autocorrelated, since the RHS
of the graph not only represents higher electro-negativity, but also a larger content of the
44
1.3 1.4 1.5 1.6
-0.2
0
0.2
0.4
0.6
-Zr
ZrV2
-ZrCr2
-ZrCr2
-ZrCr2
Zr3Fe
Zr2Fe
ZrFe2
ZrSn3
Zr5Sn3
H_sol=0
"Average" electronegativity of IM (Pauling)

E
s
o
l

(
e
V
)
Figure 4.1: E
sol
is plotted with respect to -Zr TET as the reference value.
The dotted line represents the value for which E
sol
= 0, i.e. when there is no
enthalpy gain in having the hydrogen as an adsorbed intersticial rather than in
its molecular (gas) form. The abscissae values are a ctitious form of electro-
negativity obtainedby smearing the individual electro-negitivities of each atom
into a single, uniform, weighted value across the entire phase.
non-zirconium element (since all alloying additions have higher electro-negativity than
Zr). Therefore, towards the RHS of the graph the properties of the intermetallic all
the properties, not only electro-negativity are bound to be more inuenced by the
additive element Furthermore, such trend would require the reference value (-Zr TET) to
be between the negative and the positive values, rather than to the far left of the graph.
4.2 Volumetric parameters
A second means of comparing defect energy across the dierent systems is by relating it
to the available volume in the crystal structure. Such quantity can be expressed either as
the packing fraction or as the atomic density. The latter is a much simplied quantity that
does not take into account the various site geometries, but it has the great advantage of
being easy to calculate and reliable. While the packing fraction is heavily dependent on
the radii of the atoms, which is relatively straightforward to measure for pure metals, but
rather arduous and arbitrary for intermetallic phases. Even though this was attempted,
45
and graph with a clear correlation was produced (as presented in the symposium
1
), it is
not reproduced here because sensitivity analysis showed that the values of the graph were
excessively susceptible to the measured metallic radii (for which the estimated error was
too large). It is believed that such graph could lead to incorrect interpretation of the data,
and was of little scientic relevance.
However, the solution energy vs. atomic density graph is presented in g. 4.2. To
analyse this graph it is useful to follow the lowest point for each series. The data show a
30 40 50 60 70
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
-Zr
-Zr
-ZrCr2
-ZrCr2
-ZrCr2
Zr3Fe
Zr2Fe
ZrFe2
Zr3Sn
Zr5Sn3
ZrV2
atoms/Vol (nm
-3
)
E
s
o
l

(
e
V
)
Figure 4.2: Solution energy E
sol
vs. atomic density. Only the most stable sites
were considered in the composition of this graph. The zero-line represents the
value for which there is no enthalpy gain (or loss) for the hydrogen atom to go
in solution.
clear depression with the minimum around 45-55 atoms/nm
3
. Note also that the well is
asymmetric, with the higher atomic density being a much steeper descend compared to the
low number density side. The presence of a minimum rather than a steady increase, seems
to suggest that H does not prefer the largest site possible, but rather that it prefers to feel
(and exert) a localised strain around it. This is perhaps not too surprising considering that
hydrogen diuses up the stress concentration gradient
25? ?
.
To have a better understanding of the relationship between space and stability of an
interstitial site, it is useful to compare the defect energy to the defect volume V(dened as
1
Year-4 research project presentation, nuclear group, 17-03-2011, RSM, Imperial College, London
46
the volume dierence of a defective cell compared to the perfect crystal). It is often found
in the literature that volume accommodation comes with an energy associated with it, so
that defects with greater V, which impose a higher strain to neighbouring atoms, have
higher defect energy. The present study has two important limitations when analysing
defect volumes: rstly it consider interstitial and not substitutional defects, therefore the
defect volume is unlikely to be negative; secondly it investigates only one defect specie in
dierent sites, rather than dierent species in the same site. This implies that, as well as
dierent defect neighbours, also dierent defect geometries will be compared against each
other. Inevitably, this will lead to large scatter in the results, as we can see from g. 4.3.
Even though it is hard to understand the behaviour of each phase individually, it is worth
noticing that most data series tend to higher energy values as at larger defect volumes
this is especially obvious for Zr
5
Sn
3
(dark orange) and -Zr (light purple). Furthermore,
there is a noticeable division around 4
3
. Before said value most points are negative
and mostly distributed within 0 and 0.7eV, whilst beyond it they tend to be positive and
large. In order to elucidate the general trend of the data, past the scatter, the graph has been
subdivided along the ordinate into six energy bands, and for each band the distribution of
points along the abscissae has been calculated. The six normalised distributions have been
plotted in areal histogram form in g. 4.4. In such a graph it becomes relatively straight-
forward that the most stable sites tendto have V values closer to 0 (slightly biased towards
the positive volume changes). Whilst defect that have a higher energy (lighter-coloured
bands) are predominantly those that required a larger volume accommodation.
The phases in which H seems to form defects with the smallest volumetric change are
Zr
3
Fe and Zr
3
Sn. However, these results provide no information about how easily does
each phase accommodate the change in volume required for the stable interstices, in fact
this parameter is most probably one of the main sources of the scatter.
47
-2 0 2 4 6 8
-0.5
0
0.5
1
1.5
-Zr
-Zr
Zr3Fe
Zr2Fe
ZrFe2
-ZrCr2
-ZrCr2
-ZrCr2
Zr3Sn
Zr5Sn3
ZrV2
V (
3
)
E
s
o
l

(
e
V
)
Figure 4.3: Energy of solution E
sol
vs. defect volume V. The individual trends
for each of the series could not be representedin a clean manner due to the large
scatter of the data.
Figure 4.4: The distribution of the lowest energy points is represented by the
darkest shade of blue, and the least favourable sites are represented by the
palest shade. The fraction of data points (y-axis) has been normalised by the
total number of points in each band.
48
4.3 Defects neighbouring atoms
Regardless of the comparative method used (electro-negativity, atomic density or defect
volumes), it is quite clear that the interstitials with the lowest energy are mostly observedin
the phases withthe largest Zr content (Zr
3
Xinparticular, with X=Sn,Fe). By analysing these
it became clear that the most stable sites in such alloys are the ones in which hydrogens
neighbours are solely Zr atoms. Furthermore, in the low Zr content alloys, such as ZrM
2
(with M=Fe,Cr,V), where by stoichiometry the majority of interstitial sites have more M-
type neighbours compared to Zr neighbours, the most stable ones are characterised by a
ration of 1:1 between Zr and M atoms.
With this premise in mind, the degree of electron sharing near the defect site was
investigated. This was achieved by plotting the iso-surfaces of the charge density for
the most stable interstitial congurations in each phase. ??
49
4.4 H diusion in -Zr
By observing the crystal structure of -Zr again, re-displayed below for convenience,
one can evince that diusion between basal planes can occur only by hopping between
tetrahedral and octahedral sites via the non-basal trigonal site (for simplicity we will call
this the T-O process). On the other hand migration along the c-axis can occur in two ways:
T-H-T From TET
_
at
1
3
,
2
3
,
1
8
_
, passing through HEX
_
at
1
3
,
2
3
, 0
_
, to the symmetrically
equivalent TET
_
at
1
3
,
2
3
,
1
8
_
. However this can occur only for half the height of unit
cell, since the HEX site appears only every (0002) plane.
O-t-O Alternatively, migration can occur from OCT
_
at
2
3
,
1
3
,
1
4
_
, via TRI
_
at
2
3
,
1
3
, 0
_
, to the
symmetrically equivalent OCT
_
at
2
3
,
1
3
,
1
4
_
. This process alone is sucient to obtain
c-axis diusion since it is not constrained by Zr-occupied sites.
Figure 4.5: Structure of -Zr, from bottom left the intersticial sites (red crosses
are): hexahedral
_
1
3
,
2
3
, 0
_
, tetrahedral
_
1
3
,
2
3
,
1
8
_
, non-basal trigonal
_
5
9
,
1
9
,
1
6
_
,
octahedral
_
2
3
,
1
3
,
1
4
_
and basal trigonal
_
2
3
,
1
3
, 0
_
.
Even though the former migration process cannot by it self contribute to diusion,
it is not necessarily the case that c-axis diusion is achieved only by O-t-O. In fact, T-
H-T can be coupled with the diusion from tetrahedral to octahedral and back to (a
dierent) tetrahedral site (via the T-Oprocess), to obtain an alternative route. The calculated
activation energies for both processes agree very well with experimental values, and the
respective energy paths were superimposed in g. 4.6 by plotting the vertical component
of the MEP on the abscissae.
50
0 0.5 1 1.5 2 2.5
0
0.1
0.2
0.3
0.4
0.5
0
2
4
6
8
10
z ()

H

(
e
V
)
TET
TET
HEX
nbT
nbT
OCT OCT

H

(
k
c
a
l
/
m
o
l
)
TRI
Figure 4.6: The enthalpy change is taken with respect to the lowest stable
energy state (the TET site). z represents the distance along the c-vector from
the starting interstitial site. Dashed lines are generated by mirroring the data.
Because the T-O and the T-H-T MEP are along the [0001] direction, this also
represents the total distance travelled by the hydrogen. However, for the initial
part of the red curve (and its symmetric counterpart), the migration route is
along the
_
3

308
_
direction, so that the abscissae value represent the projections
of such migration along the vertical axis.
The activation energy for the red route is dictated by the TET to OCT migration process,
which is also the mechanism trough which non-vertical diusion occurs. Therefore,
isotropic diusion can occur by this process alone. However the dierence of energy
barrier of such process compared to the blue route is minimal (0.40 compared to 0.42eV),
consequently it is safe to state that hydrogen will probably diuse using both processes. It
is worth noticing that Domains work
50
predictedthe T-O process for non-vertical diusion
(in agreement with the current work) but the O-t-O one for diusion in the other directions,
even though his values were (0.35 compared to 0.41eV).
further work: NEB on IM, ZrNi system, ZrSn2.
51
Bibliography
[1] J. Peterson, M. MacDonell, L. Haroun, and F. Monette, Radiological and Chemical
Fact Sheets to Support Health Risk Analyses for Contaminated Areas, 2007.
http://www.evs.anl.gov/pub/doc/ANL_ContaminantFactSheets_All_070418.pdf.
[2] J. R. Lamarsh and A. J. Baratta, Introduction to Nuclear Engineering (3rd Edition).
Prentice Hall, 2001.
http://www.amazon.com/Introduction-Nuclear-Engineering-John-Lamarsh/dp/0201824981.
[3] W. Yeniscavich, R. A. Wolfe, and R. M. Lieberman, Hydrogen absorption by nickel
enriched zircaloy-2, Journal of Nuclear Materials 1 (Oct., 1959) 271280.
[4] S. Glasstone and A. Sesonske, Nuclear Reactor Engineering: Reactor systems engineering.
Springer, 1994. http://books.google.com/books?id=yugQKddO82IC&pgis=1.
[5] D. R. Lide, Handbook of Chemistry and Physics, 90th edition. CRC, 90 ed., 2010.
http://www.hbcpnetbase.com/.
[6] K. Linga Murty and I. Charit, Texture development and anisotropic deformation of
zircaloys, Progress in Nuclear Energy 48 (May, 2006) 325359.
[7] L. M. Fox, P. B. Petrie, and K. Lucius, Standard Composition Library, 1998.
[8] J. H. Schemel, , A. C. B.-. o. R. Alloys, and R. Metals, ASTM manual on zirconium and
hafnium. ASTM International, 1977.
http://books.google.com/books?id=dI_LssydVeYC&pgis=1.
[9] J. Kearns, Terminal solubility and partitioning of hydrogen in the alpha phase of
zirconium, Zircaloy-2 and Zircaloy-4, Journal of Nuclear Materials 22 (1967) 11.
[10] Allegheny Technologies, Reactor Grade Zirconium Alloys for Nuclear Waste
Disposal - Technical data sheet, 2003.
[11] Y. Hatano, R. Hitaka, M. Sugisaki, and M. Hayashi, Inuence of size distribution of
Zr(Fe, Cr)2 precipitates on hydrogen transport through oxide lm of Zircaloy-4,
Journal of Nuclear Materials 248 (Sept., 1997) 311314.
52
[12] Foster, US5112573.pdf, 1992.
[13] D. Q. Peng, X. D. Bai, F. Pan, H. Sun, and B. S. Chen, Inuence of Implanted
Aluminum Ions on the Oxidation Behavior of M5 Alloy at 500C,
Oxidation of Metals 65 (Sept., 2006) 377390.
[14] C. Coleman and D. Hardie, The hydrogen embrittlement of -zirconiumA review,
Journal of the Less Common Metals 11 (Sept., 1966) 168185.
[15] C. Ells, Hydride precipitates in zirconium alloys (A review),
Journal of Nuclear Materials 28 (Nov., 1968) 129151.
[16] R. S. Daum, S. Majumdar, D. W. Bates, A. T. Motta, D. A. Koss5, and M. C. Billone,
On the Embrittlement of Zircaloy-4 Under RIA-Relevant Conditions, Zirconium in
the Nuclear Industry: Thirteenth International Symposium, Annecy, France, June 10-14,
2001, ASTM Special Technical Publication 1423. (2001) 131.
[17] E. Zuzek, J. P. Abriata, A. San-Martin, and F. D. Manchester, The H-Zr
(hydrogen-zirconium) system,
Bulletin of Alloy Phase Diagrams 11 (Aug., 1990) 385395.
[18] H. Okamoto, H-Zr (Hydrogen-Zirconium),
Journal of Phase Equilibria and Diusion 27 (Oct., 2006) 548549.
[19] N. Dupin, A thermodynamic database for zirconium alloys,
Journal of Nuclear Materials 275 (Nov., 1999) 287295.
[20] D. Setoyama and S. Yamanaka, Phase diagram of ZrOH ternary system,
Journal of Alloys and Compounds 370 (May, 2004) 144148.
[21] W. Wang and D. R. Olander, Thermodynamics of the ZrH System,
Journal of the American Ceramic Society 78 (Dec., 1995) 33233328.
[22] S. Lumley, Imperial College London, and UK, Transfer Report - intermediate draft,
2010.
[23] M. P. Puls, DEVELOPMENTS IN DELAYED HYDRIDE CRACKING IN, in
International Conference on Fracture 11. 2005.
[24] M. P. Puls, Review of the thermodynamic basis for models of delayed hydride
cracking rate in zirconium alloys,
Journal of Nuclear Materials 393 (Sept., 2009) 350367.
[25] G. McRae, C. Coleman, and B. Leitch, The rst step for delayed hydride cracking in
zirconium alloys, Journal of Nuclear Materials 396 (Jan., 2010) 130143.
53
[26] M. P. Puls, Comments on authors reply to Review of the thermodynamic basis for
models of delayed hydride cracking rate in zirconium alloys, M.P. Puls in J. Nucl.
Mater. 393 (2009) 350367, Journal of Nuclear Materials 399 (Apr., 2010) 248258.
[27] Y. S. Kim, Authors 2nd reply to comments on authors reply to Review of the
thermodynamic basis for models of delayed hydride cracking rate in zirconium
alloys, M.P. Puls in J. Nucl. Mater. 393 (2009) 350367,
Journal of Nuclear Materials 399 (Apr., 2010) 259265.
[28] Y. S. Kim, Authors reply to Review of the thermodynamic basis for models of
delayed hydride cracking rate in zirconium alloys, M.P. Puls in J. Nucl. Mater. 393
(2009) 350367, Journal of Nuclear Materials 399 (Apr., 2010) 240247.
[29] Y. S. Kim, Comments on the DuttonPuls model: Temperature and yield stress
dependences of crack growth rate in zirconium alloys,
Materials Science and Engineering: A 527 (Nov., 2010) 74807483.
[30] Y. S. Kim, Comment on The rst step for delayed hydride cracking in zirconium
alloys by G.A. McRae et al., J. Nucl. Mater. 396 (2010) 130143,
Journal of Nuclear Materials 396 (Jan., 2010) 144148.
[31] Y. Gou, Y. Li, Y. Liu, H. Chen, and S. Ying, Evaluation of a delayed hydride cracking
in Zr2.5Nb CANDU and RBMK pressure tubes,
Materials & Design 30 (Apr., 2009) 12311235.
[32] IAEA, Delayed hydride cracking in zirconium alloys in pressure tube nuclear
reactors, 2004. ww-pub.iaea.org/MTCD/publications/PDF/te_1410_web.pdf.
[33] Y. S. Kim, Precipitation of reoriented hydrides and textural change of -zirconium
grains during delayed hydride cracking of Zr2.5%Nb pressure tube,
Journal of Nuclear Materials 297 (Sept., 2001) 292302.
[34] E. Tenckho, Deformation mechanisms, texture, and anisotropy in zirconium and zircaloy.
ASTM International, 1988.
http://books.google.com/books?hl=en&lr=&id=gPc-ACTtg5kC&pgis=1.
[35] H. Sims, Radiation chemistry in the nuclear fuel cycle, in National Nuclear
Laboratory, guest Lecturer at Imperial College. 2009.
[36] G. Leli` evre, C. Tessier, X. Iltis, B. Berthier, and F. Lefebvre, Impact of intermetallic
precipitates on hydrogen distribution in the oxide layers formed on zirconium alloys
in a steam atmosphere: a 2D(3He,p)[alpha] nuclear analysis study in microbeam
mode, Journal of Alloys and Compounds 268 (Mar., 1998) 308317.
54
[37] M. S. Khan, M. S. Islam, and D. R. Bates, Cation doping and oxygen diusion in
zirconia: a combined atomistic simulation and molecular dynamics study,
Journal of Materials Chemistry 8 (1998) 22992307.
[38] D. Khatamian and F. D. Manchester, AN ION BEAM STUDY OF HYDROGEN AND
Zr-Nb ( 2 . 5 wt %) I . Diusion parameters for dense oxide DIFFUSION, Journal of
Nuclear Materials 166 (1989) 300306.
[39] B. Cox, A mechanism for the hydrogen uptake process in zirconium alloys,
Journal of Nuclear Materials 264 (Jan., 1999) 283294.
[40] Y. Hatano, K. Isobe, R. Hitaka, and M. Sugisaki, Role of Intermetallic Precipitates in
Hydrogen Uptake of Zircaloy-2, Journal of Nuclear Science and Technology 33 (1996)
944949.
[41] P. Chemelle, D. B. Knorr, J. B. Van Der Sande, and R. M. Pelloux, Morphology and
composition of second phase particles in zircaloy-2, Journal of Nuclear Materials 113
(Jan., 1983) 5864.
[42] N. Bangaru, An investigation of the microstructures of heat-treated Zircaloy-4,
Journal of Nuclear Materials 131 (Apr., 1985) 280290.
[43] C. Li, Determination of Fe and Cr content in -Zr solid solution of Zircaloy-4 with
dierent heat-treated states, Journal of Nuclear Materials 304 (Aug., 2002) 134138.
[44] T. Y. Yang, G. P. Yu, and L. J. Chen, Precipitate solubility in [beta]-quenched
Zircaloy-4 irradiated by nitrogen-ions at 470 C, Journal of Nuclear Materials 150 (1987)
6777.
[45] W. Yang, R. Tucker, B. Cheng, and R. Adamson, Precipitates in zircaloy:
Identication and the eects of irradiation and thermal treatment,
Journal of Nuclear Materials 138 (Apr., 1986) 185195.
[46] R. Kuwae, K. Sato, E. Higashinakagawa, J. Kawashima, and S. Nakamura,
Mechanism of zircaloy nodular corrosion,
Journal of Nuclear Materials 119 (Dec., 1983) 229239.
[47] R. Jerlerud P erez, C. Toolon-Masclet, J.-M. Joubert, and B. Sundman, The ZrSn
binary system: New experimental results and thermodynamic assessment, Sept.,
2008. http://oatao.univ-toulouse.fr/2322/.
[48] E. Vitikainen and P. Nenonen, Transmission electron microscopy studies on
intermetallics in some zirconium alloys,
Journal of Nuclear Materials 78 (Dec., 1978) 362373.
55
[49] G. E. Moore, Cramming more components onto integrated circuits, Electronics 38
(1965) .
[50] C. Domain, R. Besson, and A. Legris, Atomic-scale Ab-initio study of the Zr-H
system: I. Bulk properties, Acta Materialia 50 (2002) 35133526.
[51] C. Domain, R. Besson, and A. Legris, Atomic-scale ab initio study of the ZrH
system: II. Interaction of H with plane defects and mechanical properties,
Acta Materialia 52 (Apr., 2004) 14951502.
[52] C. Domain, Ab initio modelling of defect properties with substitutional and
interstitials elements in steels and Zr alloys, Journal of Nuclear Materials 351 (June,
2006) 119.
[53] IUPAC Compendium of Chemical Terminology 2006, 2006.
http://old.iupac.org/goldbook/AT06989.pdf.
[54] J. Slater, The Theory of Complex Spectra, Physical Review 34 (Nov., 1929) 12931322.
[55] P. Hohenberg and W. Kohn, Inhomogeneous Electron Gas,
Physical Review 136 (Nov., 1964) B864B871.
[56] J. Phillips, Energy-Band Interpolation Scheme Based on a Pseudopotential,
Physical Review 112 (Nov., 1958) 685695.
[57] M. L. Cohen and V. Heine, Solid State Physics Volume 24, vol. 24 of Solid State Physics.
Elsevier, 1970. http://dx.doi.org/10.1016/S0081-1947(08)60070-3.
[58] A. P. Patel, Point defects in ceramics for detector applications. PhD thesis, 2010.
[59] M. C. Payne, T. A. Arias, and J. D. Joannopoulos, Iterative minimization techniques
for ab initio total-energy calculations: molecular dynamics and conjugate gradients,
Reviews of Modern Physics 64 (Oct., 1992) 10451097.
[60] G. Kresse, M. Marsman, and J. Furthmuller, Vienna Ab-initio Simulation Package - VASP
the GUIDE. Computational Physics, Faculty of Physics, Universitat Wien, Wien,
Austra, 2009. http://cms.mpi.univie.ac.at/VASP/.
[61] QuantumWise, Atomistix Toolkit 2008.10, 2008.
http://www.quantumwise.com/documents/manuals/ATK-2008.10/chap.relax.html#fig.relax.neb1
[62] J. Goldak, L. Lloyd, and C. Barrett, Lattice Parameters, Thermal Expansions, and
Gr uneisen Coecients of Zirconium, 4.2 to 1130K,
Physical Review 144 (Apr., 1966) 478484.
56
[63] J. W. Edwards, R. Speiser, and H. L. Johnston, High Temperature Structure and
Thermal Expansion of Some Metals as Determined by X-Ray Diraction Data. I.
Platinum, Tantalum, Niobium, and Molybdenum,
Journal of Applied Physics 22 (Apr., 1951) 424.
[64] J. Thewlis and A. R. Davey, Thermal Expansion of Grey Tin,
Nature 174 (Nov., 1954) 10111011.
[65] F. H. Spedding, a. H. Daane, and K. W. Herrmann, The crystal structures and lattice
parameters of high-purity scandium, yttrium and the rare earth metals,
Acta Crystallographica 9 (July, 1956) 559563.
[66] G. Leli` evre, D. Fruchart, P. Convertc, and F. Lef` evre-Joud, Characterisation by
neutron diraction in high temperature pressurised water of the surface corrosion
and hydrogen embrittlement of zircaloy-4,
Journal of Alloys and Compounds 347 (Dec., 2002) 288294.
[67] J. Vandersande and A. Bement, An investigation of second phase particles in
Zircaloy-4 alloys, Journal of Nuclear Materials 52 (Aug., 1974) 115118.
[68] R. VERSACI and M. IPOHORSKI, Composition of Zr(Cr, Fe)2-type precipitates in
Zircaloy-2 and Zircaloy-4, Journal of Nuclear Materials 116 (June, 1983) 321323.
[69] V. Krasevec, Transmission electron microscopy study of second phase particles in
Zircaloy-2, Journal of Nuclear Materials 98 (May, 1981) 235237.
[70] D. Shaltiel, I. Jacob, and D. Davidov, Hydrogen absorption and desorption
properties of AB2 laves-phase pseudobinary compounds, May, 1977.
http://dx.doi.org/10.1016/0022-5088(77)90162-X.
[71] M. Lototsky, V. Yartys, Y. Klochko, V. Borisko, R. Starovoitov, V. Azhazha, and
P. Vyugov, Applications of ZrV hydrogen getters in vacuum-plasma devices:
Phase-structural and hydrogen sorption characteristics,
Journal of Alloys and Compounds 404-406 (Dec., 2005) 724727.
[72] F. Stein, G. Sautho, and M. Palm, Experimental determination of intermetallic
phases, phase equilibria, and invariant reaction temperatures in the Fe-Zr system,
Journal of Phase Equilibria 23 (Dec., 2002) 480494.
[73] R. B. Russel, Coecients ot thermal expansion for zirconium, Transactions of the
Metallurgical Society of Aime 200 (1954) 10451052.
[74] S. Cho, Eect of Fe addition on hydrogen storage characteristics of
Ti0.16Zr0.05Cr0.22V0.57 alloy,
Journal of Alloys and Compounds 307 (July, 2000) 304310.
57
[75] G. B. Skinner and H. L. Johnston, Thermal Expansion of Zirconium between 298K
and 1600K, The Journal of Chemical Physics 21 (Dec., 1953) 1383.
[76] J. ARDISSON, R. MANSUR, and E. DASILVA, A study of structural and electronic
properties of the alloy systems (Zr1xTix)2Fe and (Zr1xTix)3Fe in the range 0 x 0.2,
Scripta Metallurgica et Materialia 25 (June, 1991) 13271331.
[77] V. Yartys, H. Fjellvag, I. Harris, A. Riabov, M. Sorby, and I. Zavaliy, Hydrogen
ordering and H-induced phase transformations in Zr-based intermetallic hydrides,
Journal of Alloys and Compounds 293-295 (Dec., 1999) 7487.
[78] P. GORRIA, Crystallisation and polymorphic transformations in Fe?Zr amorphous
alloys obtained by high-energy ball milling,
Physica B: Condensed Matter 350 (July, 2004) E1075E1077.
[79] T. ZOTOV, E. MOVLAEV, S. MITROKHIN, and V. VERBETSKY, Interaction in
(Ti,Sc)Fe2H2 and (Zr,Sc)Fe2H2 systems,
Journal of Alloys and Compounds 459 (July, 2008) 220224.
[80] J. Mestnik Filho, X-ray and time dierential perturbed angular correlation
measurements in ZrCr2 and ZrCr2H3 Laves phase compounds,
Journal of Alloys and Compounds 224 (June, 1995) 6065.
[81] J. Soubeyroux, M. Bououdina, D. Fruchart, and L. Pontonnier, Phase stability and
neutron diraction studies of Laves phases Zr(Cr1xMx)2 with M= Mn, Fe, Co, Ni,
Cu and 0 x 0.2 and their hydrides,
Journal of Alloys and Compounds 219 (Mar., 1995) 4854.
[82] R. GONZALEZ and L. GRIBAUDO, Analysis of controversial zones of the ZrCr
equilibrium diagram, Journal of Nuclear Materials 342 (June, 2005) 1419.
[83] G. Gunnar and S. Andersson, The Crystal Structure of Zr5Sn3 and Zr3Sn, Acta
Chemica Scandinavica 14 (1960) 956957.
[84] W. Ro teutscher and K. Schubert, No Title, Zeitschrift fuer Metallkunde 56 (1965)
813.
[85] Y. U. Kwon and J. D. Corbett, The zirconium-tin system, with particular attention to
the Zr5Sn3-Zr5Sn4 region and Zr4Sn, Chemistry of Materials 2 (Jan., 1990) 2733.
[86] H. L. Luo, E. Vielhaber, and E. Corenzwit, Occurrence of A-15 Phases,
Zeitschrift fr Physik 230 (Oct., 1970) 443448.
[87] W. D aumer, H. Khan, and K. L uders, Electronic structure investigated by NMR and
superconductivity of cubic Laves-phase hydrides: V {2}HfH {x} (0x4.5) and
V {2}Hf {0.5}Zr {0.5}H {x} (0x4.8), Physical Review B 38 (Sept., 1988) 44274436.
58
[88] A. Lawson, Low-temperature crystal structures and superconductivity of
(Hf {1-x}Zr {x})V {2} alloys, Physical Review B 17 (Feb., 1978) 11361138.
[89] A. SAWATZKY, THE DIFFUSION AND SOLUBILITY OF HYDROGEN IN THE
ALPHA-PHASE OF ZIRCALOY-2, Journal of Nuclear Materials 2 (1960) 6268.
[90] J. Kearns, Diusion coecient of hydrogen in alpha zirconium, Zircaloy-2 and
Zircaloy-4, Journal of Nuclear Materials 43 (June, 1972) 330338.
59