Hydrides with the perovskite structure: General bonding and stability considerations and the new representative CaNiH3

Article abstract of DOI:10.1016/j.jssc.2005.08.026

The stability and electronic structure of perovskite hydrides ABH ₃ were investigated by means of first-principles density functional calculations. Two types of perovskite hydrides are distinguished: (1) When A and B are alkali and alkaline earth metals, the hydrides are ionic compounds with calculated band gaps of around 2 eV and higher. Their stability trend follows basically the concept of Goldschmidt's tolerance factor. (2) When A is one of the heavier alkaline earth metals (Ca, Sr, Ba) and B a transition metal, stable compounds ABH₃ result only when B is from the Fe, Co, or Ni groups. This stability trend is basically determined by effects associated with d band filling of both the transition metal and the hydride. In contrast to group (1) perovskites, the transition metal-containing compounds are metals. The synthesis of CaNiH₃ and its structure determination from CaNiD₃ is reported. This compound is a type (2) perovskite hydride with a fully occupied hydrogen position (CaNiD₃: a=3.551(4) A, d_Ni-D=1. 776(2) A). Its stability is discussed with respect to transition metal hydrides with complex anions (e.g., Mg₂NiH₄, Na ₂PdH₂, Sr₂PdH₄).

Full text of DOI:10.1016/j.jssc.2005.08.026

ARTICLE IN PRESS
Journal of Solid State Chemistry 178 (2005) 3381–3388
www.elsevier.com/locate/jssc
Hydrides with the perovskite structure: General bonding and stability
considerations and the new representative CaNiH3
Ã
Toyoto Satoa, Dag Noreusa, Hiroyuki Takeshitab, Ulrich Haussermannc,
aStructural Chemistry, Stockholm University, 10691 Stockholm, Sweden
bDepartment of Materials Science and Engineering, Faculty of Engineering, Kansai University, Yamate-cho, Suita, Osaka 564-8680, Japan
cInorganic Chemistry, Stockholm University, 10691 Stockholm, Sweden
Received 27 May 2005; received in revised form 16 August 2005; accepted 28 August 2005
Available online 3 October 2005
Abstract
The stability and electronic structure of perovskite hydrides ABH3 were investigated by means of first-principles density functional
calculations. Two types of perovskite hydrides are distinguished: (1) When A and B are alkali and alkaline earth metals, the hydrides are
ionic compounds with calculated band gaps of around 2 eV and higher. Their stability trend follows basically the concept of
Goldschmidt’s tolerance factor. (2) When A is one of the heavier alkaline earth metals (Ca, Sr, Ba) and B a transition metal, stable
compounds ABH3 result only when B is from the Fe, Co, or Ni groups. This stability trend is basically determined by effects associated
with d band filling of both the transition metal and the hydride. In contrast to group (1) perovskites, the transition metal-containing
compounds are metals. The synthesis of CaNiH3 and its structure determination from CaNiD3 is reported. This compound is a type (2)
perovskite hydride with a fully occupied hydrogen position (CaNiD3: a ¼ 3.551(4) A, dNi–D ¼ 1.776(2) A). Its stability is discussed with
respect to transition metal hydrides with complex anions (e.g., Mg2NiH4, Na2PdH2, Sr2PdH4).
r 2005 Elsevier Inc. All rights reserved.
Keywords: Perovskite structure; Metal hydrides; First-principles calculations
1. Introduction
Besides chalcogenides and halogenides, there are a few
examples of perovskite hydrides (X ¼ H). In the case of A
and B being a monovalent alkali metal (M) and a div alent
alkaline earth metal (Ae), the ternary metal hydrides have a
saline bonding character. Such hydrides are of interest for
hydrogen storage at high temperature. There are five reported
Perovskites ABX3 represent an enormously prolific family
of inorganic compounds. They constitute of two differently
sized metal cations A and B and anions X from the
chalcogen or halogen groups. Ideal perovskites adopt the
¯
cubic space group Pm3m, where X anions centre the cell
¯
hydrides with the ideal Pm3m perovskite structure: CsCaH3,
edges at 3d (1/2,0,0) and A- and B-type cations are located at
the cell centre and corners, respectively (Fig. 1). Thus,
B-type cations are octahedrally and A-type cations are
cuboctahedrally coordinated by X. Usually, SrTiO3 is
considered as the archetype of the ideal perovskite structure.
However, most perovskites appear distorted and crystallise
in a lower symmetric space group. This gives rise to diverse
magnetic and electric properties, which make perovskites
interesting and important for technological applications [1].
RbCaH3, KMgH3, BaLiH3, and SrLiH3 [2–5]. The latter are
sometimes referred to as inverse perovskites since the singly
charged M cation occupies the octahedral voids. Addition-
ally, it has been proposed that the light metal hydride LiBeH3
adopts the cubic perovskite structure and displays high-
temperature superconductivity [6]. However, the structure
and properties of LiBeH3 have been questioned [7].
Interestingly, there is a second group of perovskite
hydrides, consisting of a divalent metal and the transition
metal Pd. These hydrides—CaPdH2, SrPdH2.7, YbNiH2.7
,
YbPdH2.7, and EuPdH2.9 [8–11]—exhibit a variable degree
of hydrogen deficiency with respect to the ideal perovskite
composition. In this work we attempt to rationalise bonding
Ã
Corresponding author.
E-mail addresses: dag@struc.su.se (D. Noreus), ulrich@inorg.su.se
(U. Haussermann).
0022-4596/$ - see front matter r 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.jssc.2005.08.026

ARTICLE IN PRESS
T. Sato et al. / Journal of Solid State Chemistry 178 (2005) 3381–3388
3382
finer grains and mixed with Ni powder in a molar ratio
Ca:Ni ¼ 1:2. The metal mixture was pressed into pellets
and the pellets were put into an aluminium oxide tube,
which was reacted in a stainless steel reactor. The reactor
was heated at 10 K/min from 293 to 898 K, then kept at
898 K for 125 min, heated again at 10 K/min to 1073 K,
kept at this temperature for 120 min and finally cooled to
room temperature. The obtained product was crushed,
further Ni powder was added, and the mixtures were again
pressed into pellets. The pellets were reacted using a similar
procedure as above, but with the final temperature set to
1123 K. The final product was a mixture of CaNi3 and
Ca2Ni7 with some CaO impurity. The mixture of Ca–Ni
alloys was deuterated at 473 K at a pressure of 40 kg/cm2
for 7 days in a stainless steel reaction tube. In order to
optimise the reacting conditions, initial tries were made
using hydrogen where reaction time, pressure and tem-
perature were varied.
Hydrided and deuterated Ca–Ni alloys were investigated
by X-ray powder diffraction with a Guinier–Hagg focusing
camera of 40 mm diameter, using monochromatised CuKa1
radiation (l ¼ 1:5405980 A). Si (a ¼ 5:430879 A at 298 K)
was added as an internal standard. The films obtained were
measured in an LS 18 film scanner [19]. The program
SCANPI [20] was used to evaluate the photographs, and
the programs TREOR [21] and PIRUM [22] were used to
index the patterns. The deuterated sample was measured by
powder neutron diffraction (l ¼ 1:47 A) from 2y ¼
4:0012139:921 in steps of Dð2yÞ ¼ 0:081 at 295 K, at
NFL, Studsvik, Sweden. The Rietveld program Fullprof
was used for profile refinement of the powder neutron
diffraction data [23].
c
A
B
X
b
a
Fig. 1. Cubic perovskite structure.
and stability of perovskite hydrides. The bonding properties
of the transition metal-containing compounds should be
rather different from that of the saline compounds.
Additionally, we report on the synthesis and structural
characterisation of CaNiH3, which represents a stoichio-
metric transition metal, perovskite hydride. This is in
contrast to CaPdH2, which allegedly consists of statistically
disordered linear complexes [PdH2]2À [8]. Such molecular
units had previously been identified in Li2PdH2 and
Na2PdH2 [12,13]. In this respect reports from Orimo et al.
[14] and Chen et al. [15] on amorphous MgNiH1.9 and
MgNiH2.2, respectively, are interesting. The radial distribu-
tion function of MgNiH1.9 obtained from the measured
neutron diffraction pattern was interpreted with a coordina-
tion of two for deuterium around nickel with an Ni–D
distance of 1.7(3) A [16]. This could indicate a possible linear
NiH2 unit, similar to what has previously been found with
palladium.
No 1:1 Ca:Ni intermetallic compound exists and the 1:1
ratio in the CaNiH3 hydride is maintained with the support
of hydride ions. The instability of the metal atom frame-
work without the support of hydrogen also seems to have
some consequences for the synthesis of CaNiH3. This
hydride could not be obtained by sintering CaH2 and
nickel powder in hydrogen up to 5 MPa using various
sintering temperatures up to 623 K. Takeshita et al. [17]
found what was assumed to be CaNiH3 by carefully
disproportioning a CaNi3 hydride. Kakuta et al. [18]
reported similar results when they reacted a CaH2–33 at%
Ni mixture in an anvil cell at 5 GPa and 1073 K. But in this
case, the deuteride prerequisite for full structural refine-
ment was not synthesised. In the present work, we prepared
CaNiH3 and CaNiD3 by reacting mixtures of CaNi3 and
Ca2Ni7 alloys with hydrogen or deuterium in a rather
elaborate scheme and the structure was obtained from
neutron diffraction data of the deuteride.
2.2. Electronic structure calculations
Total energy calculations for perovskite hydrides ABH3
were performed in the framework of the frozen core all-
electron Projected Augmented Wave method [24], as
implemented in the program VASP [25]. The energy cut-
off was set to 500 eV. Exchange and correlation effects
were treated by the generalised gradient approximation,
usually referred to as PW91 [26]. The integration over the
Brillouin zone was done on special k-points determined
according to the Monkhorst–Pack scheme [27]. All
necessary convergence tests were performed and total
energies were converged to at least 1 meV/atom. The
equilibrium volume Veq and the corresponding energy Eeq
were obtained by fitting E vs. V values to a Birch–Murna-
ghan equation of state. The theoretical equilibrium lattice
parameter and some physical properties of the experimen-
tally known compounds are collected in Table 1. With the
exception of SrLiH3, the deviation between experimental
and theoretical lattice parameter is 0.5% or lower. In order
to assess zero Kelvin energies of formation for ABH3, the
same procedure was applied for hydrides AeH2 (Ae ¼ alka-
line earth metal) and MH (M ¼ alkali metal). MH was
considered in the cubic NaCl structure, and AeH2 in the
2. Experimental
2.1. Synthesis and characterisation of CaNiH3
Commercially pure granular Ca and Ni powders were
used as starting materials. The Ca granules were cut into

ARTICLE IN PRESS
T. Sato et al. / Journal of Solid State Chemistry 178 (2005) 3381–3388
3383
Table 1
Calculated lattice parameters (a), bulk moduli (B0) and band gaps (eg) of
the investigated hydrides
MAeH3
Be
1.0
0.5
a (A)
B0 (GPa)
eg (eV)
CsCaH3
RbCaH3
KMgH3
BaLiH3
SrLiH3
4.6169 (4.609)
4.5326 (4.547)
4.0181 (4.018)
4.0148 (4.023)
3.8044 (3.833)
3.5262 (3.544)
3.8490 (3.843)
3.7273
23.5
24.3
34.5
38.2
42.8
93.2
82.3
90.8
79.9
3.17
3.32
2.59
2.28
1.93
—
Ca
0.0
CaNiH3
SrPdH3
CaPdH3
CaPdH2
—
—
Mg
Cs
-0.5
3.6997 (3.69)
—
Li
Na
K
Rb
M
(a)
Experimental values for a are given in parentheses.
K
1.5
1.0
tetragonal rutile (Ae ¼ Be, Mg) and orthorhombic BaCl2-
type structure (Ae ¼ Ca, Sr, Ba).
AeMH3
3. Results and discussion
Na
0.5
3.1. Saline perovskites MAeH3 and (anti-perovskites)
AeMH3
0.0
Li
We calculated energies for the formation reactions
-0.5
Be
Mg
Ca
Sr
Ba
Ae
(b)
MH þ AeH2 ! MAeH3
ðM ¼ Li; Na; K; Rb; Cs; Ae ¼ Be; Mg; CaÞ
and
Fig. 2. Energies of formation with respect to AeH2 and MH for MAeH3
(Ae ¼ Be, Mg, Ca) (a) and AeMH3 (M ¼ Li, Na, K) (b).
AeH2 þ MH ! AeMH3
these compounds, however, our calculations yield slightly
positive formation energies [31]. Turning to anti-perovs-
kites AeMH3, only BaLiH3 and SrLiH3 yield negative
energies of formation. Indeed, these two compounds adopt
this structure type [5].
ðAe ¼ Be; Mg; Ca; Sr; Ba; M ¼ Li; Na; KÞ:
The results are depicted in Figs. 2a and b. Concerning
perovskites MAeH3, we obtain negative formation energies
for MMgH3 with M ¼ Na, K, Rb, Cs. This agrees well
with the experimental observations. The formation energy
The stability of the ideal cubic perovskite structure is
usually discussed using the Goldschmidt tolerance factor
¯
for NaMgH3 in the Pm3m perovskite structure is only
p
slightly negative. Indeed, NaMgH3 crystallises in an
orthorhombically distorted perovskite structure (i.e., the
GdFeO3 type) [28]. We calculated that the stabilisation
over the ideal perovskite structure is about 0.09 eV/Z.
KMgH3 has the most negative formation energy in the
series of Mg compounds and crystallises with the ideal
perovskite structure [4]. The ideal perovskite formation
energy of RbMgH3 is somewhat higher than that of
KMgH3. This compound adopts the hexagonal BaTiO3
structure, which is a 6H stacking variant of the hexagonal
perovskite structure [29]. As for NaMgH3, the formation
energy for CsMgH3 is only slightly negative. This indicates
that a compound with this composition should be stable,
but unlikely to crystallise in the ideal perovskite structure.
Experimentally, it has been found that CsMgH3 has a low-
and a high-pressure form [30]. The structures of both
modifications are not related to the ideal perovskite
structure. Also, CsCaH3 and RbCaH3 are reported to
t ¼ ðrX þ rAÞ=½ 2ðrX þ rBÞꢀ. This factor becomes unity if
rX ¼ rA, i.e., X and A form an ideal cubic close packing of
spheres and B fits the octahedral voids perfectly ðrB ¼
p
employed, i.e., purely ionic bonding is assumed. Empiri-
ð
2 À 1ÞrAðrX ÞÞ [1]. For the calculation of t ionic radii are
¯
cally, it has been observed that the Pm3m structure occurs
for t between 0.9 and 1 (for the archetype SrTiO3
t ¼ 1:002). Fig. 3 compiles t for MAeH3 and AeMH3
assuming a radius of 1.3 A for HÀ and using the Shannon
radii for 12-coordinated M+ and six-coordinated Ae2+
and vice versa for the anti-perovskites. As can be seen, the t
¯
values for the systems realising the Pm3m structure fall in
the range between 0.92 and 1.01, which is in agreement
with the general observation. For each series MAeH3
(Ae ¼ Be, Mg, Ca) and AeMH3 (M ¼ Li, Na, K), the
system with a t value closest to unity also corresponds to
the most stable with respect to the decomposition into
binary hydrides. This justifies the ionic bonding description
of this class of perovskite hydrides.
¯
adopt the ideal cubic Pm3m perovskite structure [2,3]. For

ARTICLE IN PRESS
T. Sato et al. / Journal of Solid State Chemistry 178 (2005) 3381–3388
3384
MAeH3
AeMH3
1.0
1.2
1.1
1.0
0.9
0.8
4
Be
0.9
Li
2
Mg
Na
0.8
0
K
Ca
-2
0.7
(a)
(b)
Na
K
Rb
Cs
Ca
Sr
Ba
-4
-6
Fig. 3. Goldschmidt tolerance factors t for MAeH3 (Ae ¼ Be, Mg, Ca) (a)
and AeMH3 (M ¼ Li, Na, K). Stable perovskites are found in the grey
areas.
Γ
M
X
Γ
R
Fig. 5. Band structure of CsCaH3.
CsCaH3
RbCaH3
their transformation behaviour at G. This splitting at G
gives rise to a characteristic two-peak shape in the DOS of
CsCaH3 below the Fermi level. The lower lying a1g band
has a dispersion of about 2 eV, and the narrower eg band of
about 0.5 eV. CsCaH3 is an insulator with a band gap of
3.17 eV. The calculated band gap in the saline perovskites
reduces with decreasing lattice parameter. For SrLiH3 with
a ¼ 3:846 A (H–H distance 2.72 A), the size of the band gap
is 1.97 eV. At the same time, the dispersion of the occupied
bands increases to about 6 eV and the two-peak shape of
the occupied states disappears. Importantly, the increasing
dispersion of the occupied bands with decreasing lattice
parameter is accompanied by an increasing admixture of
states from the small and polarising cations Mg or Li. This
is essential for the stabilisation of this class of hydride
compounds. Previously, Orgaz and Gupta analysed the
electronic structure of SrLiH3 and BaLiH3 [32]. Their
results are very similar to ours.
KMgH3
BaLiH3
3.2. The new representative CaNiH3
The X-ray diffraction pattern of hydrided Ca–Ni alloys
could be indexed with a cubic unit cell a ¼ 3:544ð1Þ A,
which is in agreement with CaNiH3 as reported by
Takeshita et al. [17] and Kakuta et al. [18]. The unit cell
of deuterated CaNiD3 was determined as a ¼ 3:5437ð8Þ A.
The observed and calculated 2y, d-values and X-ray
intensities are shown in Table 2, indicating a CsCl-type
structure for the metal atoms in CaNiH3.
SrLiH3
E-EF (eV)
The refinement of the neutron diffraction pattern was
made assuming the perovskite structure for CaNiD3 as
shown in Table 3. The Ni–D distance is 1.776(2) A and the
D–D distance 2.511(2) A. Observed and calculated inten-
sities from the Rietveld refinement are shown in Fig. 6.
Three impurity phases were identified: Ni, CaNi5 and CaO.
All were included in the refinement. The refined weight
fractions of the participating phases were CaNiD3 51(2)%,
Ni 32(2)%, CaNi5 9(1)% and CaO 8(1)%. Ni is thus the
Fig. 4. Density of states for the ideal perovskites CsCaH3, RbCaH3,
KMgH3, BaLiH3 and SrLiH3 at their theoretical equilibrium volume.
As expected, the electronic structure of the saline
perovskite hydrides is very similar. Fig. 4 shows the
electronic density of states (DOS) of the stable perovskite
hydrides. Additionally, in Fig. 5 the band structure of
CsCaH3 is depicted. We denote the three occupied bands
arising from the hydrogen states as a1g and eg according to

ARTICLE IN PRESS
T. Sato et al. / Journal of Solid State Chemistry 178 (2005) 3381–3388
3385
Table 2
Observed and calculated 2y, d-values and X-ray intensities for CaNiD3
agreement with a pure alloy not containing hydrogen. The
very similar unit cell axis values for CaNiD3 and Ni lead to
problems with the overlapping peaks in the refinement. The
Rietveld refined cell parameters were 3.538 and 3.528 A,
respectively. These values are, however, less accurate than
the values obtained from the X-ray diffraction using an
internal standard to correct the 2y scale of the patterns.
The peak width parameters could not be refined simulta-
neously, but were estimated from the diffraction pattern
and kept fixed during the refinements. The major misfits in
the diffraction pattern also occur where the diffraction
peaks of CaNiD3 and Ni overlap. Fortunately, both
structures are very simple with no positional parameters
to refine. Refining the occupancy of the deuterium atom
position led to a slightly higher (within one standard
deviation) occupancy than one. We thus assume that
CaNiD3 can be described with an ideal ABH3 perovskite
structure type having a hydrogen capacity corresponding
to 2.95 wt%. This is also in agreement with the earlier
performed volumetric determination of the hydrogen
content [17], but is in contrast to YbNiH2.7, CaPdH2,
YbPdH2.7 and EuPdH2.9 displaying the same kind of CsCl-
type metal arrangement.
The band structure and DOS of CaNiH3 are shown in
Figs. 7 and 8, respectively. The small lattice parameter of
3.5262 A implies a wide dispersion of the hydrogen-based
bands. In CaNiH3, the a1g–eg splitting of these bands is
pronounced. The lower lying a1g band is dispersed over a
region from À10 to À6 eV below the Fermi level. This band
contains a stabilising contribution from the Ni-s state.
Compared to the a1g band, the eg band is rather narrow
and confined between À6.5 and À5.5 eV. Due to a
substantial admixture of Ni eg d states, this band is
essentially responsible for Ni–H bonding. At À2 eV the
basically dispersionless, non-bonding, Ni t2g d states are
h k l
2yobs
2ycal
dobs
dcal
Iobs
Ical
1 0 0
1 1 0
1 1 1
2 0 0
2 1 1
2 2 0
25.043
35.788
44.421
51.471
64.059
75.627
25.059
35.733
44.142
51.471
64.200
75.701
3.5529
2.5070
2.0378
1.7740
1.4524
1.2554
3.5507
2.5108
2.0500
1.7753
1.4496
1.2551
2
100
o1
16
1
100
o1
14
23
6
28
7
M(6) ¼ 46.
Table 3
¯
Refinement results for CaNiD3 (space group Pm3m (no. 221) and Z ¼ 1)
Atom
Site
x
y
z
Biso
Occupancy
Ni
Ca
D
1a
1b
3d
0
0
0
0.2(2) 1.0
0.6(5) 1.0
2.4(5) 1.0
1/2
1/2
1/2
0
1/2
0
RB
RF
Wt. fraction
(%)
(%)
(%)
CaNiD3
Ni
4.40
4.13
19.3
11.5
2.47
1.97
13.6
52(2)
32(2)
9(1)
CaNi5
CaO
14.0
8(1)
Rwp ¼ 20.1%.
4000
3500
3000
2500
2000
1500
Yobs
Ycalc
1000
500
0
2
Yobs -Ycalc
60 70
-500
eg*
10
20
30
40
50
80
90
2θ (°)
0
Fig. 6. Rietveld profile refinement of powder neutron diffraction data (o)
¯
for CaNiD3 in space group Pm3m (no. 221). Reflection markers are for
CaNiD3, Ni, CaNi5 and CaO (from above), respectively.
-2
t2g
-4
eg
dominating impurity. Both the CaNiD3 and Ni phases
were of a poor crystal quality as reflected in the rather
broad diffraction lines. This is understandable because
during the hydrogenation reaction, which is performed
at rather low temperatures, a major rearrangement of
the metal atoms takes place, e.g., CaNi3+3/2H2 ¼
CaNiH3+Ni. The decomposition reaction CaNiH3 ¼
CaH2+Ni occurs at 748 K in an H2 atmosphere [17],
which prevents the employment of heat treatments for
improving crystal quality. The unit cell dimension of
CaNi5 was refined to a ¼ 4:958 and c ¼ 3:932 A in close
-6
-8
-10
-12
a1g
Γ
M
X
Γ
R
Fig. 7. Band structure of CaNiH3 at the theoretical equilibrium volume.

ARTICLE IN PRESS
T. Sato et al. / Journal of Solid State Chemistry 178 (2005) 3381–3388
3386
0.0
4d
-1.0
Ca+H2 = CaH2
H-s
-2.0
-3.0
5d
Cr Mn Fe Co Ni Cu
Ni-d
Ti
V
Fig. 9. Energy of formation with respect to the elements for CaTH3
(T ¼ Ti–Cu, Rh–Ag, Ir–Au).
Ca-s,p,d
E-EF (eV)
equilibrium lattice parameter for CaPdH3 is a ¼ 3:727 A
(cf. Table 1). When applying the virtual crystal approx-
imation for describing randomly H-deficient CaPdH2, the
lattice parameter reduces to 3.700 A, which is close to the
experimental value (a ¼ 3:69 A) [34]. On the other hand,
the calculated equilibrium lattice parameter for SrPdH3
and the experimental value for SrPdH2.7 are in good
agreement (cf. Table 1). Thus, our calculations corroborate
hydrogen deficiency for the Ca compound. We return to
this issue at the end of this chapter.
Fig. 8. Total density of states of CaNiH3 and its decomposition into H-s,
Ni-d and Ca-s,p,d contributions (from above).
centred. At even higher energies, from À1.5 to 1.5 eV, the
Ni–H antibonding eg band is distributed. This band is only
half filled, which makes CaNiH3 a metallic conductor. A
completely filled eꢁg band would put the Fermi level at the
pseudo-gap at 1.5 eV. The contribution of Ca states to the
bands below the Fermi level is small. This suggests
considering CaNiH3 as consisting of a polyanionic frame-
work [NiH3]2À counterbalanced by Ca2+ cations. Thus,
bonding in metallic CaNiH3 appears rather different
compared to the saline, insulating, perovskites. This is
also indicated in the considerably higher bulk modulus
(lower compressibility) of CaNiH3 compared to that of the
saline pervoskite hydrides.
It is interesting to speculate on the general possibility to
obtain transition metal-containing hydrogen perovskites
AeTH3. Fig. 9 gives a summary of the formation energies
Ca þ T þ 3=2H2 ! CaTH3
ðT ¼ Ti2Cu; Rh2Ag; Ir2AuÞ:
In these calculations, all components were treated as
nonmagnetic. The shape of the Ca-curve is a result of the
different stability trends of T and CaTH3 upon d band
filling: The maximum for CaCrH3 appears as a conse-
quence of the high stability (large energy of cohesion) of
elemental Cr with a half-filled d band. Perovskites CaTH3
are most stable for T ¼ Mn. This becomes clear from
Fig. 10 where the DOS of CaTH3 (T ¼ Ti–Cu) at the
computed equilibrium volumes are compiled. For T ¼ Ti
and V the nonbonding t2g and antibonding eg band are
separated by a small gap which turns into a pseudo-gap for
T ¼ Cr and Mn. For T ¼ Mn the Fermi level is exactly at
this pseudo-gap, i.e., a situation is entered where the t2g
band is completely filled and the eg band empty. Starting
off T ¼ Fe the t2g and eg bands merge and instead a
separation of the eꢁg and the high-lying Ca-based bands
appear. For CaCuH3, the egꢁ band is completely filled. The
minimum in the formation energy curve at T ¼ Ni is a
consequence of the more rapidly increasing destabilisation
of T with increasing d band filling compared to CaTH3.
Taking into account the formation energy of CaH2
(indicated in Fig. 9) perovskite hydrides are only stable
for T ¼ Fe, Co, Ni. The trend in the formation energy is
3.3. Perovskites AeTH3
Hitherto transition metal hydride perovskites AeTH3
have been found with T ¼ Ni, Pd and divalent Ae ¼ Ca,
Sr, Yb, Eu. The reported variable hydrogen deficiency of
most of these compounds appears suspicious. Apart from
CaPdH2 their hydrogen content is close to the stoichio-
metric value of 3 (i.e., in a range of 2.7–2.9; as a matter of
fact we cannot rule out such a slight hydrogen deficiency
for CaNiH3 either). The electronic structure of Pd
perovskite hydrides has been previously investigated by
Orgaz et al. [33] by employing the augmented plane wave
method. In this work, hydrogen deficiency in CaPdH2 was
simulated by taking a weighed average of the structure
factor over the three H sites, whereas for the other
compounds a stoichiometric composition was assumed.
The DOS for SrPdH3 is similar to that of CaNiH3 (cf.
Fig. 8). The main difference is that the gap between the eg
and t2g band is not present for SrPdH3 due to the lower
lying Pd-d states. The large degree of H deficiency of
CaPdH2 is remarkable. The computationally obtained

ARTICLE IN PRESS
T. Sato et al. / Journal of Solid State Chemistry 178 (2005) 3381–3388
3387
structure trend was found for hypothetical MgTH3
compounds as recently reported by Vegge et al. [35].
The stability of transition metal perovskite hydrides with
polyanionic assemblies [TH3]2À (T ¼ Ni, Pd) has especially
to be considered with respect to more hydrogen-rich
transition metal hydride compounds containing isolated
homoleptic complexes THnmÀ with formally zero-valent T.
Examples are Mg2NiH4 and Sr2PdH4 with tetrahedral
[TH4]4À [36,37], NaBaH3 with trigonal planar [TH3]3À [38]
and Na2PdH2 with linear [TH2]2À [13]. Nonstoichiometric
perovskite hydrides—if existing—could actually consist of a
mixture of polyanionic [TH3]2À and linear [TH2]2À entities.
For CaPdH2 it has been proposed that the perovskite
structure may be a consequence of linear PdH22À hydrido
complexes randomly oriented along the cubic unit cell edges
[8]. However, a careful reinvestigation of these systems has
to be performed in order to corroborate this idea.
Ti
V
Cr
Mn
Fe
Acknowledgment
Financial support has been obtained from the HYS-
TORY project under the European Commission 5FP EESD
program and by the Swedish Research Council (VR).
Hakan Rundlof, NFL, Studsvik, Sweden, is acknowledged
for collecting the powder neutron diffraction data.
References
[1] R.H. Mitchell, Perovskites: Modern and Ancient, Almaz Press Inc.,
Thunder Bay, Ont., Canada, 2002.
[2] H.H. Park, M. Pezat, B. Darriet, Rev. Chim. Min. 23 (1986) 323.
[3] H.H. Park, M. Pezat, B. Darriet, C.R. Acad. Sc. 306 (1988) 963.
[4] R. Schuhmacher, A. Weiss, J. Less-Common Met. 163 (1990) 179.
[5] C.E. Messer, J.C. Eastman, R.G. Mers, A.J. Maeland, Inorg. Chem.
3 (1964) 776.
Co
[6] A.W. Overhauser, Phys. Rev. B 35 (1987) 411.
[7] J.P. Bastide, Solid State Commun. 74 (1990) 355.
[8] W. Bronger, K. Jansen, P. Muller, J. Less-Common Met. 161 (1990)
299.
Ni
[9] W. Bronger, G. Ridder, J. Alloys Compd. 210 (1994) 53.
[10] K. Ensslen, E. Bucher, H. Oesterreicher, J. Less-Common Met. 92
(1983) 343.
[11] K.H.J. Buschow, R.L. Cohen, K.W. West, J. Appl. Phys. 48 (1977)
5289.
[12] K. Kadir, D. Noreus, Z. Phys. Chem. NF 163 (1989) 231.
[13] D. Noreus, K.W. Tornroos, A. Borje, T. Szabo, W. Bronger, H. Spittank,
G. Auffermann, P. Muller, J. Less-Common Met. 139 (1988) 233.
[14] S. Orimo, K. Ikeda, H. Fujii, Y. Fujikawa, Y. Kitano, K. Yamamoto,
Acta Mater. 45 (1997) 2271.
Cu
[15] J. Chen, T. Sakai, N. Kitamura, H.T. Takeshita, N. Kuriyama,
J. Am. Chem. Soc. 123 (2001) 6193.
E-EF (eV)
[16] S. Orimo, K. Ikeda, H. Fujii, S. Saruki, T. Fukunaga, A. Zuttel,
L. Schlapbach, Acta Mater. 46 (1998) 4519.
Fig. 10. Density of states for the ideal perovskites CaTH3 (T ¼ Ti–Cu) at
[17] H. Takeshita, T. Oishi, N. Kuriyama, J. Alloys Compd. 333 (2002) 266.
[18] H. Kakuta, T. Miyashita, Y. Goto, A. Kamegawa, H. Takamura,
M. Okada, Mater. Trans. 42 (2001) 443.
their theoretical equilibrium volume.
[19] K.E. Johansson, T. Palm, P.-E. Werner, J. Phys. E 13 (1980) 1289.
[20] P.-E. Werner, L. Eriksson, S. Salome, SCANPI, A Program for
Evaluating Guinier Photographs, Stockholm University, Stockholm,
1980.
similar for the T ¼ 4d, 5d and/or Ae ¼ Sr, Ba: Transition
metal perovskite hydrides are most stable with respect to
the decomposition into the elements when T is a group 9 or
10 metal. Interestingly, a similar stability and electronic
[21] P.-E. Werner, L. Eriksson, M. Westdahl, J. Appl Crystallogr. 18
(1985) 367.

ARTICLE IN PRESS
T. Sato et al. / Journal of Solid State Chemistry 178 (2005) 3381–3388
3388
[22] P.-E. Werner, Ark. Kemi 31 (1969) 513.
[32] E. Orgaz, M. Gupta, J. Alloys Compd. 209 (1994) 159;
E. Orgaz, M. Gupta, J. Alloys Compd. 231 (1995) 147.
[33] E. Orgaz, V. Mazel, M. Gupta, Phys. Rev. B 54 (1996) 16124.
[34] The so-called virtual-crystal approximation (VCA) is the simplest
approximation in the hierarchy of mean-field approaches for the
description of random disorder [J.S. Faulkner, Prog. Mater. Sci. 273
(1981) 3; F. Ducastelle, Order and Phase Stability in Alloys, North-
Holland, Amsterdam, 1991]. The H position was treated as a
fictitious atom with the nuclear charge 2/3 and a corresponding
number of valence electrons as to force charge neutrality.
[35] T. Vegge, L.S. Hedegaard-Jensen, J. Bonde, T.R. Munter, J.K. Norskov,
J. Alloys Compd. 386 (2005) 1.
[23] J. Rodriguez-Carvajal, FULLPROF Version 3.2 March 2005, ILL
(unpublished).
[24] P.E. Blochl, Phys. Rev. B 50 (1994) 17953;
G. Kresse, J. Joubert, Phys. Rev. B 59 (1999) 1758.
[25] G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558;
G. Kresse, J. Furthmuller, Phys. Rev. B 54 (1996) 11169.
[26] J.P. Perdew, Y. Wang, Phys. Rev. B 45 (1992) 13244.
[27] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188.
[28] E. Ronnebro, D. Noreus, K. Kadir, A. Reiser, B. Bogdanovic,
J. Alloys Compd. 299 (2000) 101.
[29] F. Gingl, T. Vogt, E. Akiba, K. Yvon, J. Alloys Compd. 282 (1999)
125.
[36] P. Zolliker, K. Yvon, P. Fischer, J. Schefer, Inorg. Chem. 24 (1985)
4177.
[30] G. Renaudin, B. Bertheville, K. Yvon, J. Alloys Compd. 353 (2003) 175.
[31] In our calculations entropy and zero point energy contributions were
not considered and it is therefore possible that compositions yielding
slightly positive formation energies adopt the perovskite structure.
[37] M.O. Martensson, M. Kritikos, D. Noreus, J. Am. Chem. Soc. 121
(1999) 10908.
[38] M. Olofsson, M. Kritikos, D. Noreus, Inorg. Chem. 37 (1998) 2900.