Intermetallics 93 (2018) 6–19

Contents lists available at ScienceDirect

Intermetallics
journal homepage: www.elsevier.com/locate/intermet

Influencing factors of atomic order in the binary sigma phase MARK

a,b a,c,∗ b d
Wei Liu , Xiao-Gang Lu , Pascal Boulet , Marie-Christine Record
a
School of Materials Science and Engineering, Shanghai University, 200444, Shanghai, PR China
b
Aix-Marseille Univ, CNRS, MADIREL, 13397, Marseille, France
c
State Key Laboratory of Advanced Special Steel, Shanghai, 200072, PR China
d
Aix-Marseille Univ, Université de Toulon, CNRS, IM2NP, Marseille, France

A R T I C L E I N F O A B S T R A C T

Keywords: The sigma phase can serve as a prototype of intermetallic phases when investigating the atomic order (i.e. atomic
Sigma phase constituent distribution or site occupancy preference on inequivalent sites of a crystal structure), as the sigma
Atomic order phase bears a broad homogeneity range and there are numerous experimental site occupancy data available for
Atomic charge the sigma phase. The present work brings new insights into the influencing factors of atomic order in the binary
Size factor
sigma phase. We calculated the atomic volumes and atomic charges of the constituent elements of 32 stoi-
Number of valence electrons
chiometric end-member compounds (i.e. the complete set of ordered configurations) of the binary sigma phase
Total number of electron shells
by using first-principles calculations. The calculation results show that besides the size factor and the number of
valence electrons, the total number of electron shells is found as the third factor that affects the atomic order of
the sigma phase. Moreover, the calculation results indicate that between the two constituent elements, the one
with larger atomic size, smaller number of valence electrons or smaller total number of electron shells prefers
occupying large coordination number (CN) sites (i.e. 4f, 8i1 and 8j) and the other prefers occupying small CN
sites (i.e. 2a and 8i2). Furthermore, we dissociated the effects of individual influencing factors on atomic order of
the sigma phase combining with the measured site occupancies from the literature.

1. Introduction
The sigma phase known as topologically close-packed (TCP) phase,
is a non-stoichiometric intermetallic compound that crystallizes in a
tetragonal structure (space group P42/mnm) with 30 atoms distributed
on five inequivalent sites, namely 2a, 4f, 8i1, 8i2 and 8j [1–3]. It is a
hard, brittle phase mostly formed between transition elements and
deteriorates various properties of many technologically important ma-
terials, such as stainless and other high-alloy steels as well as Ni-based
superalloys [1,2]. The sigma phase is reported existing in 49 different
binary alloy systems, five of which are not confirmed (i.e. Cr-Ni, Ta-V,
Ti-Mn, Zr-Ir and Zr-Re). It has attracted great technological and theo-
retical study interests since the 1950s [4].
The atomic order, i.e. atomic constituent distribution or site occu-
pancy preference on inequivalent sites of a crystal structure, is one of
the research focuses. There are extensive studies of atomic order in
intermetallic and solid state compounds, e.g. mu phase [5], sigma phase
[2], chi phase [6], L12 phase [7] and B2 phase [8]. As an important
crystallographic property, the atomic order is closely related to ther-
modynamic and thermo-physical properties of intermetallic com-
pounds, such as formation enthalpy, volume and elastic modulus [9].
The sigma phase can serve as a prototype of intermetallic phases when
investigating the atomic order, as the sigma phase bears a broad
homogeneity range and there are abundant experimental site occu-
pancy data available for the sigma phase. Numerous investigations on
the atomic order of the sigma phase have been conducted, e.g. Refs.
[1,2,10–22], by using XRD (X-ray diffraction), ND (neutron diffraction),
mathematical methods or first-principles calculations.
At first, Kasper and Waterstrat [11] indicated that for the sigma
phase, 2a and 8i2 sites should be occupied by elements to the right of
Mn in the periodic table; 4f site by elements to the left and 8i1 and 8j
sites by a mixture of both. Spooner and Wilson [13] stated that the size
of the constituent atoms of the sigma phase was a major factor in
governing the filling of 2a, 4f, and 8i2 sites but in addition some valence
electron factor governed the filling of 8i1 and 8j sites. Afterwards, a
systematic review of the sigma phase was conducted by Hall and Algie
[1] and then by Joubert [2]. A more thorough argument about the
atomic order of the sigma phase was proposed. They indicated that
atoms with large size or poor in d-electrons preferentially occupy sites
with large coordination number (CN), namely 4f, 8i1 and 8j; atoms with
small size or rich in d-electrons preferentially occupy sites with small
CN, namely 2a and 8i2.

∗
Corresponding author. School of Materials Science and Engineering, Shanghai University, 200444, Shanghai, PR China.
E-mail addresses: tianziben@163.com (W. Liu), xglu@t.shu.edu.cn (X.-G. Lu), pascal.boulet@univ-amu.fr (P. Boulet), m-c.record@univ-amu.fr (M.-C. Record).

https://doi.org/10.1016/j.intermet.2017.11.006
Received 31 August 2017; Received in revised form 5 November 2017; Accepted 7 November 2017
0966-9795/ © 2017 Elsevier Ltd. All rights reserved.
W. Liu et al. Intermetallics 93 (2018) 6–19

However, all the conclusions above, focusing on the size factor and different factors have similar effects [19] (e.g. for Cr-Fe system, larger
the number of valence electrons, cannot explain the measured site oc- size and also poorer in d-electrons of Cr comparing to Fe both cause Cr
cupancies or atomic order satisfactorily. Taking Re-Mn system as an to preferentially occupy sites with large CN, namely 4f, 8i1 and 8j).
example, the numbers of valence electrons of Re (5d56s2) and Mn In the present work, besides the size factor and the number of va-
(3d54s2) are the same so the size factor should be the primary influ- lence electrons, we also considered the total number of electron shells
encing factor on the atomic order based on the argument from Hall and as an influencing factor affecting the atomic order of the sigma com-
Algie [1] and Joubert [2]. However, large differences can be observed pounds. Finally, we dissociated the effects of individual influencing
in the site occupancies measured for sites with the same CN (8i1 vs 8j, factors on atomic order. To facilitate explanation, all the binary sigma
CN = 14) and thus with similar size, as discussed in the following compounds investigated are designated as A-B where the size of atom A
Section 4.2. In fact, the experimental site occupancies show that the is always larger than that of atom B. The sizes of the constituent ele-
atomic order among the large CN sites (i.e. 4f, 8i1 and 8j) or between ments (A and B) are determined by the cell volumes of their pure ele-
the small CN sites (i.e. 2a and 8i2) can be very large, more examples ments in the sigma phase structure from the first-principles calculations
being Nb-Re (4f vs 8i1/8j) [2], Re-Fe (8i1 vs 8j) [2]. The early conclu- [10,23]. Besides, the site occupancy of atom A is given in the present
sions [1,2] about the atomic order cannot explain the occupancy pre- work. Atom B complements atom A on different sites. We use YS(A) to
ference among the large CN sites or the small CN sites. denote the site occupancy of atom A on site S. The site occupancy of
Later on, Sluiter et al. calculated the site occupancies of the sigma atom B on site S, YS(B), is equal to 1- YS(A). The sequence of atomic
compounds in Cr-Fe [19], Os-Cr [22], Re-Ta [20], Re-W [20,21] and order of atom A was given in the present work. For example, the se-
Ru-Cr [22] systems by using a first-principle statistical thermodynamic quence of atomic order, 4f(A) > 8i1(A) > 8j(A) > 8i2(A) > 2a(A),
approach. Their calculation results do not fit the experimental data very denotes a descending site occupancy preference of atom A from the site
well, especially for Cr-Fe [19] and Os-Cr [22] systems because of their 4f to 2a. The sequence of atomic order of atom B is opposite. Moreover,
simplified assumptions. They confirmed the influencing factors on five different colours are assigned to the five different crystal sites,
atomic order of Ru-Cr and Os-Cr systems from the perspective of the namely 2a (red), 4f (blue), 8i1 (orange or magenta), 8i2 (black) and 8j
crystal structure [22]. They considered 1) the size factor (the size dif- (green), and they are used consistently throughout the present work.
ference among the five inequivalent sites), 2) electronic degeneracy
arguments and 3) the pairwise interaction (A-A, A-B or B-B) within the
2. Crystal structure of sigma phase
sigma unit cell. It was indicated that their conclusion has no predictive
value [22]. Actually, their argument about the pairwise interaction was
The sigma phase crystallizes in a tetragonal structure (space group
based on the deduction that all five crystal sites were equivalent as Cr
P42/mnm) with 30 atoms distributed on five inequivalent sites, namely
and Ru/Os do not have an overwhelming preference for any of the sites.
2a, 4f, 8i1, 8i2 and 8j. Fig. 1 presents the crystal structure of the sigma
However, this deduction is not totally correct, which will be discussed
phase, which is stacked by coordination polyhedrons of CN 12, 14 and
in Section 4.2.3.
15. Coordination polyhedrons around five inequivalent sites of the
Recently, Crivello et al. [10] systematically discussed the site occu-
crystal structure are clearly clarified in Fig. 2 and Table 1 lists the
pancy behavior of the sigma phase in binary Re-X (X = Co, Cr, Fe, Hf, Mn,
nearest neighbor (NN) sites for the five inequivalent sites. More details
Mo, Nb, Ni, Os, Ru, Ta, Tc, Ti, V, W and Zr) systems by using the com-
about the crystal chemistry of the sigma phase can be found in Ref. [2].
pound energy formalism (CEF) combined with first-principles calculations.
These studied systems include stable sigma compounds (Re-V, Re-Cr, Re-
Mn, Re-Fe, Nb-Re, Mo-Re, Ta-Re and W-Re) as well as some hypothetically 3. Methodology and calculation details
unstable ones (Ti-Re, Zr-Re, Hf-Re, Re-Co, Re-Ni, Re-Tc, Re-Ru and Re-Os).
The site occupancies of the above binary sigma compounds were calcu- 3.1. First-principles calculations
lated at 2000 K (see Fig. 9 in Ref. [10]) and Crivello et al. [10] proposed
that the site occupancy behavior of the binary sigma phase is related to First-principles calculations were performed using the plane wave
whether Re bears a larger size than the other constituent element or not. If method with projector augmented wave (PAW) pseudo-potentials [24],
Re bears a larger size (e.g. for Re-V, Re-Cr, Re-Mn, Re-Fe, Re-Co, Re-Ni, Re- as implemented in the Vienna ab initio simulation package (VASP)
Tc, Re-Ru and Re-Os systems), the site occupancy preference for Re atom [25]. We used the exchange-correlation functional within the
follows 4f > 8i1 > 8j > 8i2 > 2a and if Re bears a smaller size (e.g.
for Nb-Re, Mo-Re, Ta-Re, W-Re, Ti-Re, Zr-Re and Hf-Re systems), the site
occupancy preference for Re atom follows 8i2/2a > 8i1/8j > 4f. How-
ever, for Re-Os, Re-Ru and Re-V systems where Re bears a larger size, the
calculation results show that the site occupancy preference for Re atom
follows 8j > 8i1 (for Re-Os and Re-Ru systems) and 8i2 > 8j (for Re-V
system) which is inconsistent with their above conclusion. On the other
hand, from the experimental perspective, for Re-V system the site occu-
pancy sequence for Re atom follows 8i2 > 2a > 8i1 > 4f > 8j (see
Section 4.2) which is almost the opposite sequence as the one proposed by
Crivello et al. [10]. Moreover, the site occupancy behavior between the
same CN (i.e. CN12: 2a vs 8i2; CN14: 8i1 vs 8j) sites was not clearly
clarified in that work.
At present, plentiful site occupancy measurements have been con-
ducted on the binary sigma compounds with high accuracy, but the
influencing factors on the atomic order of the sigma compounds are still
not clear. Apparently, the size factor and the number of valence elec-
trons are not enough for the explanation of the atomic order. Another
obstacle to sort out the connections is that the experimental site occu-
pancies are results of competition and cooperation of the influencing
factors on atomic order, and it is difficult to dissociate the effects of Fig. 1. Crystal structure of the sigma phase. Atoms occupying on different Wyckoff po-
sitions, namely 2a, 4f, 8i1, 8i2 and 8j, are indicated by different fillings.
individual influencing factors, as sometimes, for a specific system

7
W. Liu et al. Intermetallics 93 (2018) 6–19

Fig. 2. Coordination polyhedrons around five in-

equivalent sites of the sigma phase structure. Atoms
occupying on different Wyckoff positions, namely 2a,
4f, 8i1, 8i2 and 8j, are indicated by different fillings.

Table 1 surfaces defining the Bader volumes [30].

Number of nearest neighbor (NN) sites for the five inequivalent sites of the sigma phase. As the Bader analysis assumes that charge density maxima are lo-
cated at the center of an atom, in the present calculations, the total
Site NN sites
charge was obtained by summing the core charge and the valence
2a 4f 8i1 8i2 8j Total charge from the VASP code and the Bader analysis was performed on
the total charge density data.
2a - 4 – 4 4 12
After the calculation, we can obtain the electronic charge and
4f 2 1 2 4 6 15
8i1 - 1 5 4 4 14 atomic volume for each atom in a unit cell of the sigma compound. To
8i2 1 2 4 1 4 12 get the atomic charge for each atom, we use the valence electrons (for
8j 1 3 4 4 2 14 which the valence states were set by VASP calculation) minus the ob-
tained electronic charge. The negative value of atomic charge indicates
that an atom tends to gain electrons and the positive one indicates that
generalized gradient approximation (GGA) as parameterized by Perdew an atom tends to lose electrons.
and Wang (i.e. GGA-PW91) [26]. The magnetic parameter for ferro-
magnetism was set and spin-polarized calculations were conducted. The 4. Results and discussion
structure was fully relaxed and after the relaxation, a final static cal-
culation was performed by adopting tetrahedron method with Blöchl In the present work, besides the size factor and the number of va-
corrections. A large plane-wave cutoff energy of 400 eV was used. The lence electrons (e), the total number of electron shells (s) is also con-
k-point meshes (8*8*15) for Brillouin zone sampling were constructed sidered to clarify the regularity of the atomic order of the sigma phase.
using the Monkhorst–Pack scheme [27]. The convergence criteria for We classify the A-B (where the size of atom A is always larger than that
electronic self-consistency and ionic relaxation were 10−5 eV/unit cell of atom B) binary sigma systems involving transition elements ac-
and 10−4 eV/unit cell, respectively. These settings ensure that the en- cording to the electron configuration of the constituent elements as
ergy converges to at least 1 meV/unit cell. To facilitate the calculations, shown in Fig. 3. The A-Al (A = Nb and Ta) systems, for which the main
ZenGen script-tool [28] was used to automatically generate input files element Al is concerned, are classified into a separate category. The
for VASP calculations. systems with experimental site occupancies available are as follows:

3.2. Bader charge analysis
Richard Bader's theory of Atoms in Molecules (AIM) [29] provides
an intuitive way of dividing molecules into atoms with definition of an
atom based purely on the electronic charge density. Zero flux surfaces
on which the charge density is a minimum perpendicular to the surface
were used to divide atoms [29]. Typically in molecular systems, the
charge density reaches a minimum between atoms which is a natural
place to separate atoms from each other [30].
The Bader charge analysis code [30] developed by the Henkelman
Group was used in the present work, which is an efficient and robust
algorithm for the Bader decomposition of charge density [31–34]. The
program can read in charge density files from the VASP code and then
output the total charge associated with each atom, and the zero flux
1) SMe (atom A bears smaller number of valence electrons than atom
B, i.e. e(A) < e(B))
V-Mn (3d34s2, 3d54s2), V-Fe (3d34s2, 3d64s2), V-Co (3d34s2, 3d74s2),
V-Ni (3d34s2, 3d84s2), Cr-Mn (3d54s1, 3d54s2), Cr-Fe (3d54s1,
3d64s2), Cr-Co (3d54s1, 3d74s2), W-Re (5d46s2, 5d56s2);
2) LAs (atom A bears larger total number of electron shells than atom
B, i.e. s(A) > s(B))
Re-Mn (5d56s2, 3d54s2);
3) LAs-LAe (atom A bears larger total number of electron shells and
larger number of valence electrons than atom B, i.e. s(A) > s(B)
and e(A) > e(B))
Ru-Cr (4d75s1, 3d54s1), Re-Cr (5d56s2, 3d54s1), Re-V (5d56s2,
3d34s2), Os-Cr (5d66s2, 3d54s1);
4) LAs-SMe (atom A bears larger total number of electron shells and

8
W. Liu et al. Intermetallics 93 (2018) 6–19

Fig. 3. A-B binary sigma systems divided into different categories according to the electron configurations of A and B. V(A) > V(B), where V(A) and V(B) are the cell volumes of A and B,
respectively, in their hypothetic sigma phase structure from the first-principles calculations [10,23]. The right angled rectangles indicate no existing sigma phase or no available
experimental site occupancies for the included systems; the rounded rectangles indicate experimental data for site occupancies are available for the included systems. Note that the
existence of the sigma phase in Cr-Ni, Ta-V, Ti-Mn, Zr-Ir and Zr-Re systems is not confirmed and these systems are labelled with italic in the figure. LAs: Atom A has larger total number of
electron shells than atom B; LAe: Atom A bears larger number of valence electrons than atom B; SMs: Atom A has smaller total number of electron shells than atom B; SMe: Atom A bears
smaller number of valence electrons than atom B.

smaller number of valence electrons than atom B, i.e. s(A) > s(B)
and e(A) < e(B))
Mo-Mn (4d55s1, 3d54s2), Mo-Fe (4d55s1, 3d64s2), Mo-Co (4d55s1,
3d74s2), Ta-Rh (5d36s2, 4d85s1), Ta-Pd (5d36s2, 4d10), W-Ru (5d46s2,
4d75s1), Re-Fe (5d56s2, 3d64s2);
5) SMs-SMe (atom A bears smaller total number of electron shells and
smaller number of valence electrons than atom B, i.e. s(A) < s(B)
and e(A) < e(B))
Nb-Re (4d45s1, 5d56s2), Nb-Os (4d45s1, 5d66s2), Nb-Ir (4d45s1,
5d76s2), Nb-Pt (4d45s1, 5d96s1), Mo-Re (4d55s1, 5d56s2), Mo-Os
(4d55s1, 5d66s2), Mo-Ir (4d55s1, 5d76s2);
6) A-Al (A = Nb and Ta)
Nb-Al (4d45s1, 3s23p1), Ta-Al (5d36s2, 3s23p1).
In the present work, 12 sigma phase systems (i.e. Cr-Mn, W-Re, Mo-
Ir, Mo-Re, Nb-Pt, Mo-Mn, Ta-Pd, Ta-Rh, Ru-Cr, Re-Mn, Nb-Al and Ta-
Al) were investigated by first-principles calculations. They are chosen
to cover all categories discussed above and considered to be typical
systems.
4.1. Size and electronic factors
4.1.1. Size difference among the five inequivalent sites
The size factor has twofold meanings. The first is the atomic size
difference between the constituent elements, which was discussed in
Section 1. The second is the size difference among crystal sites. Both CN
and nearest neighbor (NN) sites affect the size of sites, which will be
discussed in this section. If only the size factor is considered, it is ob-
vious that for the two constituent elements, the one with a large atomic
size preferentially occupies sites with large size.
In the following discussion, as shown in Figs. 4–5, the atomic charge
indicates the tendency of electron loss or gain. When the value is ne-
gative (or positive), it means the atom tends to gain (or lose) the cor-
responding number of electrons.
To clarify the size difference among the five inequivalent sites of the
sigma phase, we calculated the atomic volume and the corresponding
atomic charge of the two constituent elements in the complete set
(25 = 32) of ordered configurations (i.e. the stoichiometric end-
member compounds in the following discussions) of the sigma phase for
different binary systems. Fig. 4 presents the calculation results for the
12 typical sigma phase systems. For systems in SMe, SMs-SMe cate-
gories and some systems (i.e. Ta-Pd and Ta-Rh systems) in LAs-SMe
category, a nearly linear relationship can be observed between atomic
volume and atomic charge for a specific element on a specific site and
the linear slopes for different sites are very close to each other; for
systems in LAs, LAs-LAe, A-Al (A = Nb and Ta) categories and a system
(i.e. Mo-Mn system) in LAs-SMe category, no evident linear relationship
can be observed as the atomic volume difference for a specific element
on a specific site can be very large even when the atomic charge is
constant due to the influence of the nearest neighbor (NN) atoms.
Fig. 5 shows the atomic volume versus atomic charge of the pure
elements in the hypothetic sigma structure. In this respect, we can
observe the size difference among the five inequivalent sites without
the effect of unlike NN atoms. If we consider the linear relationship
between atomic volume and atomic charge as observed in Fig. 4, and
extrapolate this evolution to a specific atomic charge (e.g. the vertical
line in Fig. 5), we can obtain the corresponding atomic volumes for all
five sites. Then we can propose from Fig. 5 that the site size sequence
follows: 4f(CN15) > 8i1(CN14) > 8j(CN14) > 8i2(CN12) > 2a(CN12). This
result agrees with the weighted average of the interatomic distances
reported for Cr-Fe sigma by Hall and Algie [1], i.e. 2.701 Å
(4f) > 2.652 Å (8i1) > 2.638 Å (8j) > 2.526 Å (8i2) > 2.508 Å (2a).
The above results agree with the analysis by considering crystal
structure. Firstly, sites with large CN (i.e. CN = 14 or 15) are expected to
bear a large size, as with more surrounding atoms the space of a site is
larger. Thus the size difference among different CN sites are expected to be
4f(CN15) > 8i1(CN14)/8j(CN14) > 8i2(CN12)/2a(CN12). Secondly, regarding
the sites with the same CN, namely CN14 (i.e. 8i1 and 8j) and CN12 (i.e.
8i2 and 2a), we should consider the effect of the NN sites. The size of the
site is expected to be larger when the surrounding small CN sites (i.e. 2a,
8i2: CN = 12) are fewer, which implies that the surrounding large CN sites
(i.e. 4f, 8i1, 8j: CN = 15, 14) are more. Table 1 shows the number of the

9
W. Liu et al. Intermetallics 93 (2018) 6–19

Fig. 4. Atomic volume with respect to atomic charge of the constituent elements in 32 (25 = 32) stoichiometric sigma end-member compounds (Mo2Mo4Mo8Mo8Mo8,
Re2Mo4Mo8Mo8Mo8 etc.) for binary systems calculated using VASP and Bader's AIM approach. Five different colours are assigned to the five inequivalent crystal sites. Every figure
contains the results of 32 stoichiometric sigma end-member compounds for a specific system and every compound contains 30 atoms. Thus, every figure contains 30 × 32 points but some
points for the like atom on the like site are overlapped, filled triangles denote the large A atoms and dots the small B atoms in the A-B binary system. (For interpretation of the references
to colour in this figure legend, the reader is referred to the web version of this article.)

10
W. Liu et al.
Fig. 4. (continued)
NN sites for the five inequivalent sites of the sigma phase. The 8i1 site is
surrounded by four CN12 sites and the 8j site is surrounded by five CN12
sites, so the size of the 8i1 site is expected to be larger than that of 8j site.
The 8i2 site is surrounded by two CN12 sites and the 2a site is surrounded
by four CN12 sites, so the size of the 8i2 site is expected to be larger than
that of 2a site. The above analysis is well consistent with the calculation
results based on Figs. 4 and 5, i.e. the site size sequence follows
4f(CN15) > 8i1(CN14) > 8j(CN14) > 8i2(CN12) > 2a(CN12).
11
Intermetallics 93 (2018) 6–19
4.1.2. Electronic factors: valence electrons and electron shells
Fig. 4 indicates that between the two constituent elements of the
sigma phase, the one with smaller total number of electron shells or
smaller number of valence electrons tends to lose electrons; the one
with larger total number of electron shells or larger number of valence
electrons tends to gain electrons. On the other hand, Fig. 5 indicates
that generally, atoms on 2a and 8i2 sites tend to gain electrons and
atoms on 4f, 8i1 and 8j sites tend to lose electrons. It is worth men-
tioning that the atoms on 8j site have a relatively larger atomic charge
W. Liu et al. Intermetallics 93 (2018) 6–19

Fig. 5. Atomic volume with respect to atomic charge of the pure elements in the hypothetic sigma phase structure calculated using VASP and Bader's AIM approach. Five different colours
are assigned to the five inequivalent crystal sites. Every figure contains 30 points but some points for the like atom on the like site are overlapped. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)
12
W. Liu et al.
Fig. 5. (continued)
range as compared to the atoms on the other sites. In some cases, atoms
on 8j site can either tend to gain or lose electrons. Thus in general,
between the two constituent elements of the sigma phase, the one with
smaller total number of electron shells or smaller number of valence
electrons prefers occupying large CN sites (4f, 8i1 and 8j) and the one
with larger total number of electron shells or larger number of valence
electrons prefers occupying small CN sites (2a and 8i2).
The above argument about the valence electron factor is consistent
with the previous investigations [1,2,35] reporting that atoms poor in
d-electrons preferentially occupy large CN sites (4f, 8i1 and 8j) and
atoms rich in d-electrons preferentially occupy small CN sites (2a and
8i2). In the present work by considering the effect of the total number of
electron shells on atomic order, we try to explain the experimental site
occupancies in more details, such as the site occupancy preference
among the large CN sites, or between the small CN sites. Besides, Berne
[35] indicated that CN12 sites show predominance for atoms with filled
13
Intermetallics 93 (2018) 6–19
or nearly filled or else empty d shells. The reason is that the CN12 sites
have approximate icosahedral symmetry causing high degeneracy of
electronic d-like levels [22,35]. Atoms with half-filled d shells would
have high density of states when occupying the CN12 site and thus
consequently occupy high CN sites, which allows a better separation of
bonding and antibonding d bands [10].
For A-Al (A = Nb and Ta) category shown in Fig. 4 (k, l), the
characteristics of atomic charge seem a little different, i.e. the tendency
of electron loss or gain for the two constituent elements is similar.
Hence for this category, the atomic order is mainly determined by the
size factor, i.e. large atoms A occupy large CN sites (4f, 8i1 and 8j) and
small atoms B occupy small CN sites (2a and 8i2). This is consistent with
the measured occupancies presented in Section 4.2.
A-Al (A = Nb and Ta) systems could be wrongly classified into the
LAs-LAe category as Nb (4d45s1) and Ta (5d36s2) bear both larger
number of valence electrons and larger total number of electron shells
W. Liu et al.
than Al (3s23p1). However, the tendency of electron loss or gain for the
two constituent elements of A-Al category is similar to that of LAs-SMe
category. It might be caused by the compromise of the LAs (i.e. atom A
bears larger total number of electron shells than atom B) and SMe (i.e.
atom A bears smaller number of valence electrons than atom B) factors.
This is consistent with the discovery by Joubert [2], who indicates that
Al behaves as if it had 9 or 10 valence electrons in the sigma phase
structure. Actually, as can be seen in Fig. 4 (f, k, l), the behavior of A-Al
(A = Nb and Ta) systems is similar to that of Mo-Mn system that be-
longs to LAs-SMe category.
4.2. Effects of electronic factors on atomic order
In this section, we aim in dissociating the effects of individual in-
fluencing factors on atomic order based on the conclusions drawn in
Section 4.1.1 (size factor) and 4.1.2 (electronic factors) and the mea-
sured site occupancy data from the literature in order to further analyze
and explain the atomic order. The experimental site occupancies result
from the competition and cooperation of these factors.
Section 4.1.2 indicates that between the two constituent elements,
the one with smaller total number of electron shells or smaller number
of valence electrons prefers occupying large CN sites (i.e. 4f, 8i1 and 8j)
and the one with larger total number of electron shells or larger number
of valence electrons prefers occupying small CN sites (i.e. 2a and 8i2).
Thus, the SMs (i.e. Atom A has smaller total number of electron shells
than atom B) and SMe factors affect the sequence of atomic order to be
4f(A)/8i1(A)/8j(A) > 8i2(A)/2a(A); the LAs and LAe (i.e. Atom A has
larger number of valence electrons than atom B) factors affect the se-
quence of atomic order to be 8i2(A)/2a(A) > 4f(A)/8i1(A)/8j(A). To reveal
the atomic order among the large CN sites (i.e. 4f, 8i1 and 8j) and be-
tween the small CN sites (i.e. 2a and 8i2) of the sigma phase, we present
in Fig. 6 the measured site occupancies from the literature category by
category, for which only recent data were collected in order to ensure
their accuracy. Fig. 7 shows the size differences of the constituent
elements of the binary sigma phase with respect to their differences
both in the number of valence electrons and the total number of elec-
tron shells. For a specific system, when the size difference is smaller and
in the meantime, the differences in both the number of valence elec-
trons and the total number of electron shells are larger as compared to
the other systems, the electronic factors affect the atomic order more
remarkably.
4.2.1. Effect of the number of valence electrons on atomic order
Fig. 6 (a) shows that in most cases for the SMe category, the site
occupancies follow Y2a(A) > Y8i2(A) and Y4f(A) > Y8j(A) > Y8i1(A).
We showed above in Section 4.1.1 that the size sequence of the crystal
site is 4f > 8i1 > 8j > 8i2 > 2a. It is thus clear that it is the
electronic factor (SMe) that affects the sequence of atomic order to be
2a(A) > 8i2(A) and 8j(A) > 8i1(A).
Besides, for all the measurements in Fig. 6 (a), among the mea-
surements of Cr-Co, Cr-Mn and V-Ni systems, there are experimental
data indicating that 4f site is not the most favorable site for atom A,
which exhibits Y8i1(A) > Y4f(A) and Y8j(A) > Y4f(A) for Cr-Co system,
Y8j(A) > Y4f(A) for Cr-Mn and V-Ni systems. It is expected to be due to
the effect of the electronic factor (SMe) since in terms of the size factor,
4f site should be the most favorable site for atom A. It can also be seen
from Fig. 7 (a) that the electronic factor can be more remarkable for
these systems than for the other ones in SMe category, as for Cr-Mn
system, the two constituent atoms bear the smallest size difference; for
V-Ni system, the two constituent atoms bear the largest difference of
valence electrons; Cr-Co system is in between. It should be mentioned
that for V-Ni system, among the three measurements by Kasper and
Waterstrat [11], only the one for the compound V69Ni31 shows mar-
ginally Y8j(A) > Y4f(A). Thus further measurement is necessary to
confirm the occupancy behavior for 4f site in the V-Ni system.
Therefore, we can expect that the SMe factor affects the atomic
14
Intermetallics 93 (2018) 6–19
order along the sequence of 8j(A) > 8i1(A) > 4f(A) > 2a(A) > 8i2(A).
Vice versa, the LAe factor affects the atomic order along the sequence of
8i2(A) > 2a(A) > 4f(A) > 8i1(A) > 8j(A).
4.2.2. Effect of the total number of electron shells on atomic order
For LAs category presented in Fig. 6 (b), measured site occupancies
for Re-Mn system are available only. The figure shows
Y8i2(A) > Y2a(A) and Y8i1(A) > Y8j(A). Moreover, the difference in A
occupancies on 8i2 and 2a sites, and on 8i1 and 8j sites for LAs category,
is too large to be explained only with the size factor. It is expected that
the LAs factor affects the atomic order in a sequence of 8i2(A) > 2a(A)
and 8i1(A) > 8j(A). By combining the discussion in the beginning of
Section 4.2 that LAs factor leads to 8i2(A)/2a(A) > 4f(A)/8i1(A)/8j(A), we
obtain 8i2(A) > 2a(A) > 8i1(A) > 8j(A).
On the other hand, both LAs-LAe and LAe categories need to be
considered to extrapolate the occupying sequence of 4f site affected by
the LAs factor, as there is no directly experimental information avail-
able. For LAs-LAe category, Fig. 7 (b) indicates that the electronic
factors affect the atomic order most remarkably for Re-V system. Fig. 6
(d) shows that for Re-V system, the site occupancies follow
Y8i2(A) > Y2a(A) > Y8i1(A) > Y4f(A) > Y8j(A) and the difference in
A occupancies on 8i1 and 8j sites is dramatically large. Besides, the first-
principles calculation results from Crivello et al. [10] also indicate the
sequence of site occupancies of Y8i1(A) > Y4f(A) > Y8j(A) among
large CN sites for Re-V system with high content of Re (see Fig. 9 in Ref.
[10]). Thus we can expect the electronic factors (both the LAs and LAe
factors) affect the atomic order in a sequence of 8i2(A) >
2a(A) > 8i1(A) > 4f(A)/8j(A) for LAs-LAe category. As discussed above,
for LAe and LAs-LAe categories the electronic factors affect the atomic
order in a sequence of 8i2(A) > 2a(A) > 4f(A) > 8i1(A) > 8j(A) and
8i2(A) > 2a(A) > 8i1(A) > 4f(A)/8j(A), respectively. In this regard, the
LAs factor must affect the atomic order in a sequence of 8i1(A) > 4f(A).
Thus the LAs factor affects the atomic order in a sequence of
8i2(A) > 2a(A) > 8i1(A) > 4f(A)/8j(A). Vice versa, the SMs factor
affects the atomic order in a sequence of 4f(A)/8j(A) > 8i1(A)
> 2a(A) > 8i2(A).
4.2.3. Experimental site occupancies
The dissociating effects of individual influencing factors on atomic
order as discovered in Section 4.1.1 (size factor), 4.2.14.2.1 (valence
electron factor) and 4.2.2 (electron shell factor) can well explain all the
accurate experimental data and first-principles calculations as will be
discussed in the following.
4.2.3.1. SMe category (Fig. 6 (a)). The influencing factors on atomic
order for this category are the size and SMe factors, which lead to a
sequence of 4f(A) > 8i1(A) > 8j(A) > 8i2(A) > 2a(A) and
8j(A) > 8i1(A) > 4f(A) > 2a(A) > 8i2(A), respectively. The evident
difference is the opposite effects of the two factors on the atomic order
of the sites with similar CN (i.e. CN12: 8i2 vs 2a, and large CN: 4f vs 8i1
vs 8j). For example, the SMe factor causes the atomic order of
2a(A) > 8i2(A) for CN12 sites while the size factor results in
8i2(A) > 2a(A). As seen in Fig. 6 (a), most experimental data show
Y2a(A) > Y8i2(A) and Y8j(A) > Y8i1(A) implying that the SMe factor
affects the atomic order on these sites more remarkably than the size
factor.
It should be mentioned that for V-Ni system, Joubert [2] also
measured the site occupancies by using the XRD method, but the data
are not accurate due to the weak XRD contrast between the two con-
stituent elements, which were not considered in the present work. On
the contrary, the early data measured by Kasper and Waterstrat [11]
are more accurate since they conducted neutron diffraction (ND) ex-
periments yielding a better resolution for site occupancy measurements.
4.2.3.2. LAs category (Fig. 6 (b)). The influencing factors on the atomic
order for this category are the size and LAs factors, which lead to a
W. Liu et al. Intermetallics 93 (2018) 6–19

Fig. 6. Site occupancy, YS (A), varies with mole fraction of A, X(A). Measured data are from Refs. [2,3,11–18,36–41]. Five different colours are assigned to the five inequivalent crystal
sites, namely 2a, 4f, 8i1, 8i2 and 8j. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

15
W. Liu et al. Intermetallics 93 (2018) 6–19

Fig. 7. Molar volume difference with respect to difference between the numbers of valence electrons of the constituent elements (A and B) of the binary sigma compounds. The difference
between the total numbers of electron shells of the elements, n, is indicated. s(A) and s(B): total numbers of electron shells of atom A and B, respectively; e(A) and e(B): numbers of valence
electrons of atom A and B, respectively; Vm(A) and Vm(B): CALPHAD assessed molar volumes of atom A and B, respectively [23]. The assessed molar volume for Mn can be found in
Appendix A.

sequence of 4f(A) > 8i1(A) > 8j(A) > 8i2(A) > 2a(A) and
8i2(A) > 2a(A) > 8i1(A) > 4f(A)/8j(A), respectively. The common
characteristic is the consistent effects of the two factors on the atomic
order of the sites with the same CN (i.e. CN12: 8i2 > 2a, and CN14:
8i1 > 8j). The experimental data indicates that the size factor
ultimately determines the site occupancy of 4f site for Re-Mn system
as presented in Fig. 6 (b).
4.2.3.3. LAs-SMe category (Fig. 6 (c)). The influencing factors on the
site occupancies for this category are the size, SMe and LAs factors,
which lead to a sequence of 4f(A) > 8i1(A) > 8j(A) > 8i2(A) > 2a(A),
8j(A) > 8i1(A) > 4f(A) > 2a(A) > 8i2(A) and 8i2(A) > 2a(A)
> 8i1(A) > 4f(A)/8j(A), respectively. The size and SMe factors cause
atom A to prefer occupying large CN sites (i.e. 4f, 8i1 and 8j) while the
LAs factor causes atom A to prefer occupying small CN sites (i.e. 2a and
8i2). All experimental data in Fig. 6 (c) show Y4f(A)/Y8i1(A)/
Y8j(A) > Y8i2(A)/Y2a(A) indicating that the combined effects of the
size and SMe factors play more important role as compared to the LAs
factor. In addition, the effects of the size and LAs factors on the atomic
order of the sites with the same CN (i.e. CN12: 8i2 vs 2a, and CN14: 8i1
vs 8j) are opposite as compared to that of the SMe factor. For example,
the size and LAs factors cause the atomic order of 8i2(A) > 2a(A) for
CN12 sites while the SMe factor results in 2a(A) > 8i2(A).
For Mo-Fe, Mo-Mn and Re-Fe systems, most measurements in Fig. 6
(c) present Y8i1(A) > Y8j(A) and marginally Y8i2(A) > Y2a(A) im-
plying the effect of the SMe factor on atomic order on these sites is less
significant than the combined effects of the LAs and size factors. Among
the three mentioned systems, for Re-Fe system, the size and valence
electrons differences between the two constituent elements are the
smallest and the electron shell difference, i.e. n = s(A)-s(B) = 2, is the
largest as shown in Fig. 7 (c). Thus the LAs factor more significantly
affects the atomic order of Re-Fe system as compared to the other two
systems, which causes the site occupancy difference between 8i1 and 8j
sites for Re-Fe system to be the largest as presented in Fig. 6 (c).
For Mo-Co system, Fig. 6 (c) shows Y8i1(A) ≈ Y8j(A) and marginally
Y8i2(A) > Y2a(A) indicating that the SMe factor affects the atomic
order on these sites almost equivalent as the combined effects of the LAs
and size factors. For Ta-Rh system, Fig. 6 (c) presents Y8i1(A) ≈ Y8j(A)

16
W. Liu et al.
and Y2a(A) > Y8i2(A) (with large difference) implying that the SMe
factor affects the atomic order more dramatically than the above
mentioned systems. Besides, the measurements for Ta-Rh system pre-
sent Y8i1(A)/Y8j(A) > Y4f(A) as shown in Fig. 6 (c). This is due to the
combining effects of the SMe and LAs factors cause the atomic order to
follow a sequence of 8j(A) > 8i1(A) > 4f(A) and 8i1(A) > 4f(A)/8j(A)
respectively. Joubert [2] also conducted the measurements for Ta-Pd
sigma compound, but the measured compositions describe a tie-line in
disagreement with the published phase diagram. Hence the data were
not considered.
For W-Ru system, the measurements in Fig. 6 (c) show that
Y2a(A) > Y8i2(A) (due to the SMe factor), Y8i1(A) > Y8j(A) (due to the
LAs and size factors) and Y8i1(A) > Y4f(A) (due to the SMe and LAs
factors).
4.2.3.4. LAs-LAe category (Fig. 6 (d)). The influencing factors on the
atomic order for this category are the size, LAs and LAe factors leading
to a sequence of 4f(A) > 8i1(A) > 8j(A) > 8i2(A) > 2a(A),
8i2(A) > 2a(A) > 8i1(A) > 4f(A)/8j(A) and 8i2(A) > 2a(A)
> 4f(A) > 8i1(A) > 8j(A), respectively. The interesting point is the
opposite effects of the size and electronic factors on the atomic order
between the sites with large and small CN sites i.e. the size factor causes
atom A to prefer occupying large CN sites (i.e. 4f, 8i1 and 8j) while the
electronic factor causes atom A to prefer occupying small CN sites (i.e.
2a and 8i2).
The site occupancy behavior for this category is quite different as
compared to the other categories which roughly follow the sequence of
a completely ordered state (i.e. large atom A occupies large CN sites;
small atom B occupies small CN sites) as defined by Joubert [2]. This
completely ordered state is defined according to the size factor. For Os-
Cr, Re-V and Ru-Cr systems, the site occupancy behavior is far deviated
from the completely ordered state indicating that the electronic factors
dramatically affect the atomic order. Among the three systems, for Re-V
system, the measurement presents a completely opposite sequence to
the ordered state. The experimental site occupancy follows
Y8i2(A) > Y2a(A) > Y8i1(A) > Y4f(A) > Y8j(A). It indicates sig-
nificant effects of the electronic factors for Re-V system which can also
be verified from Fig. 7 (b). As for Re-Cr system, a less deviation from the
ordered state can be seen in Fig. 6 (c) implying that the size factor plays
a significant role.
As mentioned in Section 1, Sluiter and Pasturel [22] clarified that
for Ru-Cr and Os-Cr systems, all five crystal sites were equivalent as
there is no overwhelming ordering preference of constituent elements.
However, according to the discussion about LAs-LAe category above,
we can conclude that the reason for no overwhelming ordering pre-
ference is due to a competition among the effects of size and electronic
factors rather than an equivalence of five crystal sites. The five crystal
sites of the sigma phase are inequivalent for all systems.
4.2.3.5. SMs-SMe category (Fig. 6 (e)). The influencing factors on the
atomic order for this category are the size, SMs and SMe factors, which
lead to a sequence of 4f(A) > 8i1(A) > 8j(A) > 8i2(A) > 2a(A), 4f(A)/
8j(A) > 8i1(A) > 2a(A) > 8i2(A) and 8j(A) > 8i1(A)
> 4f(A) > 2a(A) > 8i2(A), respectively. The main characteristic for
this category is that both the size and electronic factors cause atom A to
prefer occupying large CN sites (i.e. 4f, 8i1 and 8j) and atom B to prefer
occupying small CN sites (i.e. 2a and 8i2).
4.2.3.6. A-Al (A = Nb and Ta) category (Fig. 6 (f)). As mentioned in
Section 4.1.2, A-Al category behaves like LAs-SMe category with regard
to the electron loss or gain. Thus the influencing factors on atomic order
for this category can be roughly considered as the size, SMe and LAs
factors which lead to a sequence of 4f(A) > 8i1(A) > 8j(A)
> 8i2(A) > 2a(A), 8j(A) > 8i1(A) > 4f(A) > 2a(A) > 8i2(A) and
8i2(A) > 2a(A) > 8i1(A) > 4f(A)/8j(A), respectively.
Fig. 6 (f) shows that generally atom A (A = Nb and Ta) and Al
17
Intermetallics 93 (2018) 6–19
occupy large CN sites (i.e. 4f, 8i1 and 8j) and small CN sites (2a and 8i2),
respectively. Similar to Ta-Rh and W-Ru systems in LAs-SMe category,
the measurements in Fig. 6 (f) present Y8i1(A)/Y8j(A) > Y4f(A).
4.3. Discussion
In view of the size and electron configuration of the constituent
elements, we classified all the binary sigma phase systems into different
categories. With this classification approach, it is possible to study the
tendency of electrons loss or gain of the constituent elements and obtain
the regularity of atomic order.
The present findings cannot only explain various results for stable
sigma phases but also the calculation results for hypothetically unstable
sigma phases. As mentioned in Section 1, Crivello et al. [10] calculated
the site occupancies of the sigma phase in binary Re-X (X = Co, Cr, Fe,
Hf, Mn, Mo, Nb, Ni, Os, Ru, Ta, Tc, Ti, V, W and Zr) systems including
stable sigma compounds (Re-V, Re-Cr, Re-Mn, Re-Fe, Nb-Re, Mo-Re, Ta-
Re and W-Re) as well as some hypothetically unstable ones (Ti-Re, Zr-
Re, Hf-Re, Re-Co, Re-Ni, Re-Tc, Re-Ru and Re-Os). Some calculation
results cannot be well explained in that work as already mentioned in
Section 1 but are interpretable by using the analysis method of the
present work as follows.
The inconsistent calculated site occupancy behavior for the same CN
sites (i.e. 2a vs 8i2; 8i1 vs 8j) is due to the different effects of the in-
fluencing factors. For example, the calculated site occupancies of Re-Os
(belonging to SMe category) system present the sequence of
8j(A) > 8i1(A) due to the more significant effect of the SMe factor than
the size factor; the calculated site occupancies of Re-Tc (belonging to
LAs category) and Re-V (belonging to LAs-LAe category) systems pre-
sent the sequence of 8i1(A) > 8j(A) due to the combined effects of the
size, LAs and LAe factors.
In addition, we are also expecting to use the present method to
investigate the atomic order for other non-stoichiometric intermetallic
compounds, like Chi phase and Mu phase, etc.
5. Conclusions
Totally 12 typical binary sigma phase systems were selected to
conduct first-principles calculations. We obtained the atomic volume
and the corresponding atomic charge of the constituent atoms for each
system in a complete set of ordered configurations of the sigma phase,
from which we can discover the tendency of electron gain or loss of the
constituent atoms and the size difference among the five inequivalent
crystal sites. It is found out that besides the size factor and the number
of valence electrons, the total number of electron shells is the third
factor that affects the atomic order of the sigma phase. The experi-
mental site occupancies are results of competition and cooperation of
the size factor and electronic factors (i.e. the number of valence elec-
trons and the total number of electron shells). The present work tries to
dissociate the effects of individual influencing factors on the atomic
order for the binary sigma phase. All accurately measured site occu-
pancies can be well explained by using the following conclusions.
1. Both the coordination number (CN) and nearest neighbor (NN)
crystal sites affect the size of the five crystal sites of the sigma phase.
The size sequence of the five inequivalent crystal sites of the sigma
phase follows 4f(CN15) > 8i1(CN14) > 8j(CN14) > 8i2(CN12)
> 2a(CN12).
2. For the two constituent elements in the binary sigma compounds,
the one with smaller total number of electron shells or smaller
number of valence electrons tends to lose electrons; the one with
larger total number of electron shells or larger number of valence
electrons tends to gain electrons. On the other hand, atoms on 2a
and 8i2 sites tend to gain electrons and those on 4f, 8i1 and 8j sites
tend to lose electrons. Hence between the two constituent elements,
the one with a larger size, smaller total number of electron shells or
W. Liu et al. Intermetallics 93 (2018) 6–19

smaller number of valence electrons prefers occupying large CN sites and electronic influencing factors.
(i.e. 4f, 8i1 and 8j) and the one with a smaller size, larger total
number of electron shells or larger number of valence electrons
prefers occupying small CN sites (i.e. 2a and 8i2). Acknowledgements
3. The dissociated effects of the electronic factors leads to the fol-
lowing atomic order: This work was supported by the Chinese National Key Project of
LAs: 8i2(A) > 2a(A) > 8i1(A) > 4f(A)/8j(A); SMs: 4f(A)/ Science and Technology (Grant number: 2012ZX06004-012). W. Liu
8j(A) > 8i1(A) > 2a(A) > 8i2(A); LAe: gratefully acknowledges the financial support from the China
8i2(A) > 2a(A) > 4f(A) > 8i1(A) > 8j(A); SMe: Scholarship Council. X.-G. Lu acknowledges the High Performance
8j(A) > 8i1(A) > 4f(A) > 2a(A) > 8i2(A). Computing Center of Shanghai University for providing the calculation
facility.
The atomic order of a specific system is a compromise of the size

Appendix A. Molar volume assessments for the sigma phase in the X-Mn (X = Cr, Mo, Re, Tc, Ti and V) binary systems

By using the CALPHAD approach, Liu et al. [23] have assessed the molar volume of the sigma phase at room temperature and atmospheric
pressure for binary systems concerning 20 elements Al, Au, Co, Cr, Fe, Ir, Mo, Nb, Ni, Os, Pd, Pt, Re, Rh, Ru, Ta, Tc, V, W and Zr. As a supplement of
that work, we assessed X-Mn (X = Cr, Mo, Re, Tc, Ti and V) sigma binary systems by using the same model (for details see Ref. [23]). Besides, the
assessed results for elements Cr, Mo, Re, Tc and V were adopted from Ref. [23] directly.
During the assessments, recent data from Joubert [2] are given more considerations. The assessed results are shown in Fig. A.1 and the one for
element Mn is 7.203 × 10−6 m3/mol. The agreement between the different data is fairly good. For Ti-Mn system, the sigma phase was reported
(Vm = 7.19 × 10−6 m3/mol at around 48.6 at.% Mn) [52] but not confirmed. The present work does not attempt to assess the volume of Ti-Mn
system, but present the first-principles value for pure Ti in the sigma phase structure in Fig. A.1.

Fig. A.1. Calculated molar volumes of X-Mn (X = Cr, Mo, Re, Tc, Ti and V) sigma phase compared with the experimental data [2,42–51]. Note that for one sigma phase, the same color is
assigned to both the calculated line and the experimental data.

References 3674–3686, http://dx.doi.org/10.1021/ic302142w.

[1] E.O. Hall, S.H. Algie, The sigma phase, Metall. Rev. 11 (1966) 61–88.
[2] J.-M. Joubert, Crystal chemistry and Calphad modeling of the σ phase, Prog. Mater.
Sci. 53 (2008) 528–583, http://dx.doi.org/10.1016/j.pmatsci.2007.04.001.
[3] J. Cieślak, M. Reissner, S.M. Dubiel, J. Wernisch, W. Steiner, Influence of compo-
sition and annealing conditions on the site-occupation in the σ-phase of Fe–Cr and
Fe–V systems, J. Alloys Compd. 460 (2008) 20–25, http://dx.doi.org/10.1016/j.
jallcom.2007.05.098.
[4] G.J. Dickins, A.M.B. Douglas, W.H. Taylor, Structure of the sigma-phase in the iron-
chromium and cobalt-chromium systems, Nature 176 (1951) 192.
[5] J.-M. Joubert, N. Dupin, Mixed site occupancies in the μ phase, Intermetallics 12
(2004) 1373–1380, http://dx.doi.org/10.1016/j.intermet.2004.04.036.
[6] J.-M. Joubert, M. Phejar, Crystal chemistry and Calphad modelling of the χ phase,
Prog. Mater. Sci. 54 (2009) 945–980, http://dx.doi.org/10.1016/j.pmatsci.2009.
04.002.
[7] M.J. Liang, Z. Chen, Y.X. Wang, J.X. Zhang, Q.B. Lai, Microscopic phase-field study
on atomic site occupation probability in L12–Ni3(Al1−xFex) structure, Mater. Lett.
62 (2008) 4332–4335, http://dx.doi.org/10.1016/j.matlet.2008.07.016.
[8] Y. Wu, W. Zhang, J. Guo, J. Hou, X. Li, R. Huang, X. Ma, Q. Zhang, The first-
principles study on the occupation behavior and the ductility mechanism of Zr in
Ni–Ni3Al system with lattice misfit, J. Mater. Sci. Technol. 30 (2014) 517–522,
http://dx.doi.org/10.1016/j.jmst.2014.03.018.
[9] W. Liu, X.-G. Lu, P. Boulet, M.-C. Record, Q.-M. Hu, Influence of atomic order on the
enthalpy of formation and bulk modulus of the sigma phase, Fluid Phase Equilibria
(2017), http://dx.doi.org/10.1016/j.fluid.2017.10.006 accepted.
[10] J.-C. Crivello, A. Breidi, J.-M. Joubert, χ and σ phases in binary rhenium–transition
metal systems: a systematic first-principles investigation, Inorg. Chem. 52 (2013)
[11] J.S. Kasper, R.M. Waterstrat, Ordering of atoms in the σ phase, Acta Crystallogr. 9
(1956) 289–295, http://dx.doi.org/10.1107/S0365110X56000802.
[12] C.G. Wilson, F.J. Spooner, Ordering of atoms in the sigma phase FeMo, Acta
Crystallogr. 16 (1963) 230–231, http://dx.doi.org/10.1107/S0365110X63000566.
[13] F.J. Spooner, C.G. Wilson, Ordering in binary σ phases, Acta Crystallogr. 17 (1964)
1533–1538, http://dx.doi.org/10.1107/S0365110X64003826.
[14] S.H. Algie, E.O. Hall, Site ordering in some σ-phase structures, Acta Crystallogr. 20
(1966) 142.
[15] R.W. Spor, H. Claus, P.A. Beck, X-Ray diffraction study of ordering in two sigma
phases, Adv. X-Ray Anal. 10 (1967) 213–220.
[16] L.M. Alte da Veiga, M. Costa, M.J.M. De Almeida, L.R. Andrade, A. Matos Beja, The
ordering of the σ phases Cr2Ru and Cr2Os, Acta Crystallogr. B 36 (1980)
1750–1757.
[17] J.M. de Freitas, L.M. Alte da Veiga, W. Gonschorek, The ordering of the sigma phase
V62Co38, Portgal Phys. 13 (1982) 113–119.
[18] H.L. Yakel, Atom distributions in sigma phases. I. Fe and Cr atom distributions in a
binary sigma phase equilibrated at 1063, 1013 and 923 K, Acta Crystallogr. Sect. B
39 (1983) 20–28, http://dx.doi.org/10.1107/S0108768183001974.
[19] M.H.F. Sluiter, K. Esfarjani, Y. Kawazoe, Site occupation reversal in the Fe-Cr σ
phase, Phys. Rev. Lett. 75 (1995) 3142–3145, http://dx.doi.org/10.1103/
PhysRevLett.75.3142.
[20] C. Berne, M. Sluiter, Y. Kawazoe, A. Pasturel, Ordering effects in the Re–W and
Re–Ta sigma phases, J. Phys. Condens. Matter 13 (2001) 9433.
[21] C. Berne, M. Sluiter, Y. Kawazoe, T. Hansen, A. Pasturel, Site occupancy in the Re-W
sigma phase, Phys. Rev. B 64 (2001), http://dx.doi.org/10.1103/PhysRevB.64.
144103.
[22] M.H.F. Sluiter, A. Pasturel, Site occupation in the Cr-Ru and Cr-Os σ phases, Phys.
Rev. B 80 (2009), http://dx.doi.org/10.1103/PhysRevB.80.134122.

18
W. Liu et al.
[23] W. Liu, X.-G. Lu, Y.-L. He, L. Li, Modeling of molar volume of the sigma phase
involving transition elements, Comput. Mater. Sci. 95 (2014) 540–550, http://dx.
doi.org/10.1016/j.commatsci.2014.08.015.
[24] G. Kresse, D. Joubert, From ultrasoft pseudopotentials to the projector augmented-
wave method, Phys. Rev. B 59 (1999) 1758.
[25] G. Kresse, J. Furthmüller, Efficient iterative schemes for ab initio total-energy cal-
culations using a plane-wave basis set, Phys. Rev. B 54 (1996) 11169–11186.
[26] J.P. Perdew, Y. Wang, Accurate and simple analytic representation of the electron-
gas correlation energy, Phys. Rev. B 45 (1992) 13244–13249, http://dx.doi.org/10.
1103/PhysRevB.45.13244.
[27] H.J. Monkhorst, J.D. Pack, Special points for Brillouin-zone integrations, Phys. Rev.
B 13 (1976) 5188–5192, http://dx.doi.org/10.1103/PhysRevB.13.5188.
[28] J.-C. Crivello, R. Souques, A. Breidi, N. Bourgeois, J.-M. Joubert, ZenGen, a tool to
generate ordered configurations for systematic first-principles calculations: the
Cr–Mo–Ni–Re system as a case study, Calphad 51 (2015) 233–240, http://dx.doi.
org/10.1016/j.calphad.2015.09.005.
[29] R.F.W. Bader, Atoms in Molecules: a Quantum Theory, Oxford University Press,
Oxford, New York, 1994.
[30] Henkelman Group, Bader Charge Analysis, http://Theory.cm.utexas.edu/
Henkelman/Code/Bader/.
[31] G. Henkelman, A. Arnaldsson, H. Jónsson, A fast and robust algorithm for Bader
decomposition of charge density, Comput. Mater. Sci. 36 (2006) 354–360, http://
dx.doi.org/10.1016/j.commatsci.2005.04.010.
[32] E. Sanville, S.D. Kenny, R. Smith, G. Henkelman, Improved grid-based algorithm for
Bader charge allocation, J. Comput. Chem. 28 (2007) 899–908, http://dx.doi.org/
10.1002/jcc.20575.
[33] W. Tang, E. Sanville, G. Henkelman, A grid-based Bader analysis algorithm without
lattice bias, J. Phys. Condens. Matter 21 (2009) 084204, , http://dx.doi.org/10.
1088/0953-8984/21/8/084204.
[34] M. Yu, D.R. Trinkle, Accurate and efficient algorithm for Bader charge integration,
J. Chem. Phys. 134 (2011) 064111.
[35] C. Berne, A. Pasturel, M. Sluiter, B. Vinet, Ab initio study of metastability in re-
fractory metal based systems, Phys. Rev. Lett. 83 (1999) 1621.
[36] J. Cieślak, S.M. Dubiel, M. Reissner, Site occupancy and lattice parameters in sigma-
phase Co-Cr alloys, Acta Crystallogr. B 68 (2012) 123–127, http://dx.doi.org/10.
1107/S0108768112006234.
[37] J. Cieślak, S.M. Dubiel, J. Przewoznik, J. Tobola, Structural and hyperfine char-
acterization of σ-phase Fe–Mo alloys, Intermetallics 31 (2012) 132–136, http://dx.
19
Intermetallics 93 (2018) 6–19
doi.org/10.1016/j.intermet.2012.06.015.
[38] J. Cieślak, S.M. Dubiel, J. Zukrowski, J. Tobola, Sigma-phase in the Fe-Re alloy
system: experimental and theoretical studies, Mater. Chem. Phys. 139 (2013)
590–595.
[39] S.A. Farzadfar, M. Levesque, M. Phejar, J.-M. Joubert, Thermodynamic assessment
of the molybdenum–rhenium system, Calphad 33 (2009) 502–510, http://dx.doi.
org/10.1016/j.calphad.2009.02.001.
[40] J.-M. Joubert, C. Pommier, E. Leroy, A. Percheron-Guégan, Hydrogen absorption
properties of topologically close-packed phases of the Nb–Ni–Al system, J. Alloys
Compd. 356–357 (2003) 442–446, http://dx.doi.org/10.1016/S0925-8388(02)
01238-0.
[41] A. Boulineau, J.-M. Joubert, R. Černý, Structural characterization of the Ta-rich part
of the Ta–Al system, J. Solid State Chem. 179 (2006) 3385–3393, http://dx.doi.org/
10.1016/j.jssc.2006.07.001.
[42] Carlile, Quoted from the Power Diffraction File (PDF), Oxford, England.
[43] W.B. Pearson, J.W. Christian, W. Hume-Rothery, New sigma-phases in binary alloys
of the transition elements of the first long period, Nature 167 (1951) 110, http://dx.
doi.org/10.1038/167110a0.
[44] B.F. Decker, R.M. Waterstrat, J.S. Kasper, Formation of sigma phase in the Mn-Mo
system, J. Met. 197 (1953) 1476.
[45] P. Greenfield, P.A. Beck, Trans. AIME 200 (1954) 252.
[46] H.P. Stüwe, Description of the sigma phase as a structure with sphere packing,
Trans. Met. Soc. AIME 215 (1959) 408–411.
[47] K.P. Gupta, N.S. Rajan, P.A. Beck, Effect of Si and Al on the stability of certain sigma
phases, Trans. Met. Soc. AIME 218 (1960) 617–624.
[48] J.B. Darby, D.J. Lam, L.J. Norton, J.W. Downey, Intermediate phases in binary
systems of technetium-99 with several transition elements, J. Common Met. 4
(1962) 558–563, http://dx.doi.org/10.1016/0022-5088(62)90044-9.
[49] N.V. Ageev, V.S. Schekhtman, The crystal chemistry of the compounds of rhenium
with transition metals, in: B. Goner (Ed.), Rhenmium, Elsevier, New York, 1962, pp.
46–53.
[50] E.M. Savitskii, M.A. Tylkina, Phase diagrams of rhenium with transition metals, in:
B. Goner (Ed.), Rhenmium, Elsevier, New York, 1962, pp. 67–83.
[51] S. Setz, H. Nowotny, F. Benesovsky, Untersuchungen in den systemen: mangan-
(Vanadin, rhenium, Eisen)-Silicium, Monatsh Chem. 99 (1968) 2004–2015.
[52] R.P. Elliott, W. Rostoker, Structure of the phase TiMn and the indexing of powder
patterns of sigma type phases, Trans. AIME 197 (1953) 1203–1204.