Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.issp.ac.ru/lhpp/PapersAntonov/172.pdf
Äàòà èçìåíåíèÿ: Thu Aug 13 16:33:25 2015
Äàòà èíäåêñèðîâàíèÿ: Sun Apr 10 03:13:29 2016
Êîäèðîâêà: ISO8859-5

Ïîèñêîâûå ñëîâà: http www.badastronomy.com phpbb index.php
Acta Materialia 98 (2015) 416-422

Contents lists available at ScienceDirect

Acta Materialia
journal h omepage: www.elsevi er.c om/locate/actamat

Structure and chemical bonding in MgNi2H3 from combined high resolution synchrotron and neutron diffraction studies and ab initio electronic structure calculations
V.A. Yartys a,b,, V.E. Antonov c,d, D. Chernyshov e, J.-C. Crivello f, R.V. Denys a, V.K. Fedotov c, M. Gupta g, V.I. Kulakov c, M. Latroche f, D. Sheptyakov h
a

Institute for Energy Technology, Kjeller, Norway Norwegian University of Science and Technology, Trondheim, Norway c Institute of Solid State Physics RAS, 142432 Chernogolovka, Moscow district, Russia d National University of Science and Technology `MISIS', 119049 Moscow, Leninskii prosp. 4, Russia e Swiss-Norwegian Beam Lines, ESRF, Grenoble, France f Institut de Chimie et des Matùriaux Paris-Est, ICMPE-CNRS-UPEC, Thiais, France g Institut de Chimie Molùculaire et des Matùriaux, ICMMO-CNRS-Universitù Paris Sud, Orsay, France h Laboratory for Neutron Scattering and Imaging, Paul Scherrer Institut, 5232 Villigen PSI, Switzerland
b

article

info

abstract
Our earlier study Yartys et al. (2015) showed that at high hydrogen pressures, hexagonal MgNi2 undergoes a hydrogen assisted phase transition into the orthorhombic MoSi2-type structure. Here we report on a combined high resolution synchrotron and neutron diffraction investigation of the crystal structure of MgNi2D3, and ab initio calculation of its electronic structure that revealed the nature of the metal-hydrogen bonding. The diffraction data (293 and 1.8 K) are well described with a Cmca unit cell with H atoms filling the deformed octahedra Mg4Ni2 and the positions within the buckled nets -Ni-H-Ni-H- penetrating through the structure. DFT and phonon calculations showed that the Cmca structure of MgNi2D3 is the most stable, both from the electronic structure and the lattice dynamical arguments. The Bader charge analysis indicates an electronic transfer from Mg (Ð1.59eÐ) to Ni (+0.21eÐ), H1 (+0.55eÐ) and H2 (+0.31eÐ). The phonon dispersion curves of MgNi2H3 show positive frequencies, indicating that the structure is mechanically stable. The calculated gross heat of formation for the Cmca phase of MgNi2H3 is Ð37.3 kJ/mol-H2, which makes it more stable by 3 kJ/mol-H2 than the prototype structures tested in Yartys et al. (2015). The stability of the Cmca crystal structure of MgNi2H3 is enhanced by the formation of the directional Ni-H covalent bonds supplemented by the electron transfer from Mg to both Ni and H. The heat capacity as a function of temperature is obtained by phonon calculation in the quasi-harmonic approximation. ã 2015 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Article history: Received 8 July 2015 Accepted 20 July 2015

Keywords: Metal hydrides Crystal structure High pressures Neutron diffraction DFT Phonon Magnesium Nickel

1. Introduction In our earlier study, we have shown that the hydrogenated MgNi2 intermetallic compound with C36 Laves-type structure forms a trihydride MgNi2H3 and undergoes a hydrogen-induced phase transformation into an orthorhombic MoSi2-type related structure [1]. The hydrogenation was carried out in a hydrogen (deuterium) gas compressed to 2.8-7.4 GPa at 300 ÀC.

Corresponding author at: Institute for Energy Technology, Kjeller, Norway.
E-mail addresses: (V.A. Yartys). volodymyr.yartys@ife.no, volodymyr.yartys@gmail.com

The orthorhombic structure of MgNi2D3 was studied using time-of-flight neutron scattering in the d-spacing range between 1 and 5 Õ performed at the Joint Institute for Nuclear Research in Dubna, Russia. These studies showed that hydrogen atoms fill two types of interstices, inside the Mg4Ni2 octahedrons and inside the buckled nickel nets [1]. It should be noted, however, that phase transformations in the MgNi2-H(D)2 system are rather complex and can result in the formation of tetragonal and orthorhombic modifications of MgNi2H(D)3 according to Kataoka and coworkers [2,3]. In this connection, a high resolution powder diffraction investigation of MgNi2D3 would be of great value in order to improve quality of the raw experimental data and determine the nature of chemical

http://dx.doi.org/10.1016/j.actamat.2015.07.053 1359-6454/ã 2015 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.


V.A. Yartys et al. / Acta Materialia 98 (2015) 416-422

417

bonding stabilizing the formed trihydride and causing the rebuilding of the metal sublattice. It is well known that intermetallic hydrides demonstrate a close interrelation between their crystal chemistry and hydrogen storage behaviours thus allowing optimization of the H storage performance [4]. Hydrogen accommodation by the metal lattice is typically accompanied by a modest (few percent) increase of the interatomic metal-metal distances, while the H atoms fill the available interstices in the intermetallic structures. However, in a number of cases a formation of the metal hydride leads to rebuilding of the metal sublattice. Such a rebuilding is of particular interest as it could be related to the unconventional metal-hydrogen bonding in the hydrides thus formed. It is exactly the case of the MgNi2H3 hydride. In this hydride, only small part of the hydrogen atoms fills the interstitial sites available in the structure of the virgin MgNi2 compound, whereas the major part of the H atoms is exclusively bound to the nickel atoms and produce ÑÑÑH-Ni-H-Ni ÑÑÑ nets because of the formation of directional, partially covalent Ni-H bonds [1]. The objective of the current study is: (a) application of combined synchrotron X-ray diffraction and neutron diffraction to the characterization of the crystal structure of MgNi2H(D)3; (b) theoretical density functional theory (DFT) and phonon calculations based on the initial high accuracy crystallographic data to describe the nature of the chemical bonding and the thermodynamic properties of the studied hydride. 2. Experimental 2.1. Synthesis of MgNi2 and MgNi2D3 The starting MgNi2 material was the same as that used in [1] and prepared from high-purity Mg (99.8%) and Ni (99.995%) powders mixed with a slight excess of Mg (2 wt.%) as compared to the stoichiometric ratio Mg:Ni = 1:2. The mixture was compacted by pressing into pellets with a diameter of 15 mm. The pellets were sintered in argon gas at 800 ÀC for 12 h and quenched into ice water. XRD study showed the formation of a MgNi2 intermetallic compound crystallizing with a Laves type hexagonal MgNi2 type structure (C36) with the unit cell parameters a = 4.82565(6), c = 15.8323(3) Õ. This compound constituted about 90 wt.% of the material, the impurities being Ni metal, MgO and MgNi3 compound [1]. A 1000 mg sample of the MgNi2-based deuteride was prepared by a 24 h exposure of the powdered MgNi2 to a D2 pressure of 2.8 GPa at 300 ÀC. The sample was prepared in 5 batches, each containing appr. 200 mg of the material. Before releasing the pressure, the synthesized deuteride was rapidly cooled (quenched) to 100 K. The hydrogenation method is described in more detail elsewhere [5]. The experiment was carried out in a Toroid-type high-pressure chamber [6] using AlD3 as an internal deuterium source. In agreement with results of [1] for the MgNi2-D sample prepared in the same way, our quenched sample consisted of the orthorhombic MgNi2D3 phase with admixture of MgO and NiD deuteride; the latter was formed from the impurities of Ni and MgNi3 contained in the starting material. When the quenched sample was further heated to room temperature, the NiD phase rapidly decomposed to Ni metal and D2 gas, while the MgNi2D3 phase did not lose deuterium for a week at least. In the present work, we examined the MgNi2-D sample composed of all 5 synthesized batches mixed together and powdered in an agate mortar in an Ar glove box. The resulting powder sample was exposed to room temperature and pressure in an Ar

atmosphere for a few days needed for the sample transportation between our institutions. Rest of the time the sample was stored in liquid nitrogen to prevent the deuterium loss and oxidation by air. 2.2. Synchrotron X-ray diffraction The study was performed at the beam station BM1A at the Swiss-Norwegian Beam Lines, European Synchrotron Research Facility, Grenoble, France, using a multipurpose diffractometer based on the PILATUS2 M detector. The data were collected at 293 K using monochromatic X-ray beam with a wavelength of 0.68894(1) Õ in the 2h range 2.5-73.2À; step size 0.021748À. The wavelength was calibrated by collecting the data for a standard sample of LaB6 (NIST standard). The measured MgNi2D3 sample was enclosed in a sealed glass capillary with a diameter of 0.3 mm. 2.3. Neutron powder diffraction (NPD) The MgNi2D3 deuteride was studied by neutron diffraction at the Spallation Neutron Source SINQ at Paul Scherrer Institute, Villigen, Switzerland, using a high resolution powder diffractometer HRPT in the high intensity mode (k = 1.494 Õ, 2h range 4.5- 164.7À, step 0.05À). The sample with the overall mass of 1 g was placed into a 6 mm vanadium sample holder. Loading the deuteride into the sample holder was performed in an argon glove box. Two data sets were measured, at 293 K and at 2 K. The latter dataset was collected on a sample placed into a standard orange cryostat. 2.4. Combined SR XRD and NPD Rietveld profile refinements Combined Rietveld refinements of the neutron and synchrotron powder diffraction data collected at 293 K and at 2 K were performed using the GSAS software [7]. 2.5. Electronic structure calculations The calculations are based on the density functional theory (DFT) and performed in the same conditions (method, accuracy) as in the previous study [1]. We use the projector augmented wave method (PAW) [8] implemented in the Vienna Ab initio Simulation Package (VASP) [9,10] with the Perdew-Burke-Ernzerhof exchange-correlation functional [11] including semi-core p electrons of Mg and Ni. After the necessary tests to control the stability of the energy differences between the phases, the energy cut-off for the PAWs was set to 800 eV. Phonon calculations are carried out in the harmonic (HA) and quasi-harmonic (QHA) approximation [12] from the supercell approach (2 Ò 1 Ò 2: 96 atoms) with the finite displacement method [13] using the Phonopy code [14]. Charge transfers are computed using Bader's prescription [15,16]. 3. Results and discussion 3.1. Thermal stability of MgNi2D
3

The thermal stability and the total hydrogen content of the sample were studied by deuterium desorption in the temperature range 15-350 ÀC in the regime of heating at a rate of 2 ÀC/min in a closed-volume, pre-evacuated measuring system. In this experiment, we used a portion of the sample with a mass of 150 mg. The initial pressure in the system was near 0 bar, and the final pressure was 2.97 bar. The results are shown in Fig. 1. As seen from Fig. 1, the total amount of deuterium released from the MgNi2-D sample reached 3.6 wt.% with a peak of


418

V.A. Yartys et al. / Acta Materialia 98 (2015) 416-422

resolved extra peaks (0 2 1, 2 2 1 and 1 3 2) suggested that the structure has a C centred orthorhombic unit cell instead of the F type cell. One of the possible space groups is Cmca (No. 64), which is related to the S.G. Fmmm via the group-subgroup relations [Fmmm (No. 69) ? Cmca (No. 64; setting 1): a ? a; b ? c; c ? b]. This group provided an excellent description of the observed diffraction pattern (see the fit of the SR XRD pattern in Fig. 2). Consequently, the Cmca space group was used in the profile analysis of the neutron powder diffraction patterns. 3.3. Neutron powder diffraction study at 293 K During the deuteration, the MgNi2 compound underwent a volumetric expansion of 15.6%, from V(MgNi2) = 319.291/8 = 39.91 Õ3/f.u. to V(MgNi2D3) = 189.17/4 = 47.29 Õ3/f.u. The NPD data collected at 293 K were refined together with the SR XRD data discussed in Section 3.2. The refinement showed excellent agreement between the experimental and calculated patterns (Fig. 3). Crystal structure data for the MgNi2D3 deuteride are given in Table 1. In the structure of MgNi2D(H)3 hydrogen atoms occupy two types of sites, (a) inside the deformed octahedra Mg4Ni2 and (b) within the buckled nets -Ni-H-Ni-H- containing infinite bended spirals penetrating through the structure. This is schematically shown in Fig. 4. As is also seen from Fig. 4, the D2 position of the deuterium atoms has chair coordination by the neighbouring Ni atoms. Fig. 5 compares the Fmmm and Cmca structures of MgNi2H(D)3. As one can see, both the Mg4Ni2 octahedra and buckled networks ÑÑÑNi-H2-Ni-H2ÑÑÑ undergo a significant deformation in the Cmca structure. This deformation makes the Cmca structure more thermodynamically favourable than Fmmm. The bonding Me-H(D) distances in the Cmca structure are: D1- Ni 1.6213(5) Õ; D1-Mg 2.2951(1) Õ; D1-Mg 2.3431(1) Õ; D2-Ni 1.7444(4) Õ; D2-Ni 1.7947(4) Õ. These distances well agree with the average Mg-H and Ni-H separations observed in the structures of binary and ternary hydrides. As an example, consider 1.97- 2.17 Õ for the Mg-H and 1.52-1.74 Õ for the Ni-H distances in La2MgNi9D13 deuteride [17]. The shortest D-D separations in the structure are in agreement with the ``rule of 2 Õ'' [4]; D2-D2: 2.295 Õ and 2.402 Õ; D1-D2: 2.536 Õ.

Fig. 1. Thermal desorption traces of deuterium release from the MgNi2D pre-exposed to ambient conditions for a few days.

2Ñ93

sample

deuterium desorption at appr. 205 ÀC. According to the synchrotron XRD investigation (see Section 3.2), the sample was composed of about 90 wt.% MgNi2D$3. Two impurity phases (total content $10 wt.%) were identified as MgO and Ni containing no deuterium. Consequently, the MgNi2D$3 deuteride contained 3.6/0.9 = 4.0 wt.% D that corresponds to the stoichiometry MgNi2D2.93. 3.2. Synchrotron X-ray diffraction study In agreement with our earlier study [1], the refinements of the SR XRD pattern concluded that MgNi2D3 crystallises with orthorhombic structure. The refinements indicated presence of two impurity phases, nickel and magnesium oxide (both coming from the original alloy). However, the refinements also showed that the Fmmm space group (S.G.) proposed in [1] does not allow one to describe all observed reflections. The strongest extra peaks are marked by arrows in Fig. 2. We have tried several other space groups with a lower symmetry, including the Pmmm space group proposed in [2,3]. These refinements did not provide satisfactory improvements. The observed extinctions of the Bragg indexes of the strongest

Fig. 2. SR XRD pattern (k = 0.68894 Õ) of MgNi2D3 at 293 K used in the combined refinements of the SR XRD and NPD patterns. Rp = 5.5%, Rwp = 5.7%, RF^2 = 3.4. Vertical bars show positions of the Bragg peaks for the phase constituents (from top to bottom): MgNi2D3 (90.9%); Ni (7.6%); MgO (1.5%). Crosses represent the experimental data, solid lines are for the calculated profile and difference plot. Inset: strongest extra peaks and their hkl indexes.

Fig. 3. NPD pattern (k = 1.494 Õ) of MgNi2D3 at 293 K used in the combined refinements of the SR XRD and NPD pattern. NPD: Rp = 3.7%, Rwp = 4.9%, RF^2 = 4.4. Vertical bars show positions of the Bragg peaks for the phase constituents (from top to bottom): MgNi2D3 (90.9%); Ni (7.6%); MgO (1.5%). Crosses represent the experimental data, solid lines are for the calculated profile and difference plot. Inset: strongest extra peaks and their hkl indexes.


V.A. Yartys et al. / Acta Materialia 98 (2015) 416-422

419

Table 1 Crystal structure data for the deuteride MgNi2D3 at 293 K from the combined refinements of the SR XRD and NPD data and from the theoretical calculations. Sp.gr. Cmca (No. 64). Experimental data: a = 4.5902(1); b = 8.7943(2); c = 4.6861(1) Õ; V = 189.17(1) Õ3. Theoretical calculations: sp.gr. Cmca (No. 64). a = 4.5356; b = 8.8407; c = 4.6938 Õ; V = 188.21 Õ3. Atom Wyckoff site Experimental results x Mg Ni D1 D2 4a 8f 4b 8e 0 0 Í
1 =

Theoretical calculations z 0 0.0363(1) 0 1/4 U
iso

y 0 0.3167(1) 0 0.2199(2)

Ò 100 (Õ2)

x 0 0 Í
1 =

y 0 0.3154 0 0.2143

z 0 0.0478 0
1 =

4

0.65(3) 0.30(1) 2.7(1) 2.7(1)

4

4

Site occupancy factors for all atoms were fixed to 1.0 and not refined. Phase fractions: MgNi2D3 - 90.9(1) wt.%; Ni (Fm3m a = 3.5265(1) Õ)-7.6(1) wt.%; MgO (Fm3m; a = 4.2135(6) Õ)-1.5(1) wt.%.

Fig. 6. NPD pattern (k = 1.494 Õ) of MgNi2D3 at 2 K. Rp = 3.8%, Rwp = 4.8%, RF^2 = 3.2. Vertical bars show positions of the Bragg peaks for the phase constituents (from top to bottom): MgNi2D3 (91.9%); Ni (6.8%); MgO (1.3%). Crosses represent the experimental data, solid lines are for the calculated profile and a difference plot.

Fig. 4. Crystal structure of MgNi2D3. Two types of sites occupied by deuterium include a Mg4Ni2 octahedron for H1(D1) and a chair Ni4 configuration for H2(D2) located within the buckled Ni-H nets containing bended spirals -Ni-H2-Ni-H2-.

small contraction of the unit cell slightly exceeding 1% in total. Interestingly, this contraction is very anisotropic and is confined to the [100] direction only. Indeed, Da/aRT = Ð1.18%; Db/bRT = Ð0.05%; Dc/cRT = Ð0.01%; DV/VRT = Ð1.25%. The NPD pattern of MgNi2D3 collected at 2 K is shown in Fig. 6. Results of the refinements of the crystal structure of MgNi2D3 at 2 K are summarized in Table 2. The bonding Me-H(D) distances in this structure show only minor changes due to the cooling from 293 to 2 K and attain the following values: D1-Ni 1.6336(9) Õ; D1- Mg 2.2680(1) Õ; D1-Mg 2.3428(1) Õ; D2-Ni 1.732(4) Õ; D2-Ni 1.8094(9) Õ. 3.5. DFT calculations of the electronic structure of MgNi2H

3

Fig. 5. Relationship between the crystal structures of MgNi2D3 in the space group Fmmm [1] and in the space group Cmca (present study).

3.4. Neutron powder diffraction study at 2 K Crystal structure data for MgNi2D3 were also collected at 2 K (neutron powder diffraction study only). This study showed that the cooling of the sample from 293 to 2 K is accompanied by a

Starting with the crystal structure data experimentally determined in the present study, the theoretical calculations allowing a full relaxation of the unit cell led to a symmetry converged to the base centred orthorhombic Cmca space group (S.G. No. 64), with very comparable values of the cell parameters and internal positions to those from the experiments, see Table 1. One may notice that the Cmca space group and the monoclinic C2/m group (S.G. No. 12) resulting from ab initio calculations [1], are both maximal non-isomorphic subgroups of the Fmmm (S.G. No. 69) symmetry proposed for the MgNi2H3 phase on the experimental basis [1]. Since the atomic environments are very similar, the electronic density of states (DOS) of MgNi2H3 calculated in the Cmca S.G. and shown in Fig. 7, is very similar to the DOS in the Fmmm S.G. (shown in Fig. 10 of Ref. [1]). Thus, the discussion is intentionally omitted and we suggest reading the Section 3.5.2 of [1] for


420

V.A. Yartys et al. / Acta Materialia 98 (2015) 416-422

Table 2 Crystal structure data for MgNi2D3 at 2 K. Sp.gr. Cmca, a = 4.5360(2), b = 8.7895(4), c = 4.6857(2) Õ; V = 186.81(2) Õ3. Atom Mg Ni D1 D2 Wyckoff site 4 8 4 8 a f b e x 0 0 1/2 1/4 y 0 0.3157(1) 0 0.2143(2) z 0 0.0456(2) 0 1/4 U
iso

Ò 100 (Õ2)

0.0(-) 0.0(-) 1.5(1) 1.6(1)

* *

Phase fractions: MgNi2D3 - 91.9(1) wt.%; Ni (Fm3m; a = 3.5160(3) Õ)-6.8(1) wt.%; MgO (Fm3m; a = 4.206(1) Õ)-1.3(1) wt.%. Site occupancy factors for all atoms were fixed to 1.0. * Atomic displacement parameters.

Fig. 7. Total electronic DOS of the Cmca phase of MgNi2H3 (left vertical scale) and the number of its electrons (right vertical scale). The Fermi level is chosen as the origin of the energies.

complementary details. The directional Ni-H covalent bonds associated with short Ni-H distances are supplemented, as in [1], by an electron transfer from Mg to H and Ni atoms. The Bader charge analysis is also similar and predicts an electronic transfer from Mg (Ð1.59eÐ) to Ni (+0.21eÐ), H1 (+0.55eÐ) and H2 (+0.31eÐ).

The large difference in the charge transfer between H1 and H2 is due to their different local environment: as one can see from Fig. 4, H1 is surrounded by 2 Ni and 4 Mg atoms, while the first nearest neighbours of H2 are Ni atoms, the Mg atoms being more distant neighbours. Concerning the MgNi2 + 3/2 H2 ? MgNi2H3 reaction, the DFT and phonon calculations of all products and reactants allowed us to obtain the energy of reaction and zero-point energy (ZPE). The gross heat of formation of the Cmca hydride, without ZPE correction, is DH = Ð37.35 kJ/mol-H2 that makes this hydride by about 3 kJ/mol more stable than all the prototype structures previously tested in [1]. The ZPE energy contribution is however larger, nevertheless the final ZPE corrected heat of formation is still the lowest in the Cmca structure with DHcorr = Ð27.89 kJ/mol-H2. The values obtained in [1] for the Fmmm structure are DH = Ð34.10 kJ/mol-H2 for the gross heat of formation and DHcorr = Ð26.15 kJ/mol-H2 for its ZPE corrected value. The phonon dispersion curves of MgNi2H3 calculated in the Cmca symmetry are presented in Fig. 8. All frequencies are positive therefore indicating that the structure is mechanically stable. An analysis of the different vibrational contributions can be obtained from the partial phonon DOS shown in Fig. 9. At lower frequencies, the contributions from the heaviest Ni and Mg atoms are dominant with some coupled modes, whereas the contributions at higher frequencies come from the hydrogen vibrations. The high-energy very localized mode at %48.5 THz (1 THz = 4.15 meV) is believed to be determined by the strong interactions of H1 with its 2 Ni first neighbours located at short distances, 1.647 Õ in the DFT relaxed structure, while the H2 optic modes located between 21.5 and 35.0 THz, are much softer since the H2 distances to its 4 Ni neighbours are much larger, ranging from 1.728 to 1.818 Õ. The high energy localized modes of H1 obtained in the present calculation have frequencies %48.5 THz, which are slightly smaller than %51 THz calculated in [18] for the low temperature phase of Mg2NiH4 and associated with very short Ni-H distances, %1.54-1.56 Õ, H being in a distorted tetrahedral environment. The lowest H2 frequencies are comparable to the experimental characteristic optic phonon frequencies %21.5 THz of NiH1.05 [19], where H occupies octahedral interstices

Fig. 8. Phonon dispersion curves of MgNi2H3 in the Cmca space group.


V.A. Yartys et al. / Acta Materialia 98 (2015) 416-422

421

Fig. 9. Partial phonon DOS of MgNi2H3 in the Cmca space group.

q with a Ni-H distance of 1.86 Õ. The strong ~ dependence of the H2 optic modes points to significant H-H long range interaction. In the 11.5-17.5 THz range, the H1 vibration modes can be mostly associated with the interaction of H1 with its 4 Mg neighbours in the plane. The calculated heat capacities of the MgNi2 and MgNi2H3 compounds at constant volume, Cv, and at constant pressure, Cp, are given in Fig. 10. As expected, the Cp values are enhanced relative to Cv, since Cp = CV + a2TV/b, where V(T) is the molar volume, a = (1/V)(oV/oT)p the thermal coefficient of volume expansion, and b = Ð(1/V)(oV/oP)T the isothermal compressibility. Fig. 11 shows the temperature dependence of the unit cell volume for these compounds calculated in QHA. One can see that the volume of the hydride increases more rapidly at high temperatures. The temperature dependence of the thermal coefficient of volume expansion a(T) is also given as a helpful information for the reader. The calculated Cp(T) of MgNi2 agrees with experimental results of Ref. [20] obtained at low and moderate temperatures. The experimental values of Cp measured at higher temperatures in Ref. [21] noticeably differ from our estimates. The discrepancy could be due, in part, to some anharmonic contributions neglected in our calculations, but further experimental confirmation is also desirable. The calculated Cp in the high temperature limit is slightly larger than the 3R/mol-atom %24.9 J/K/mol-atom of the Dulong- Petit law. At low and moderate temperatures, when the high frequency H vibrations are not yet excited, the heat capacity of the hydride MgNi2H3 on a per atom basis should be lower than that of MgNi2, as found in Fig. 10. As a conclusion from the theoretical part of the paper, the Cmca structure of MgNi2H3 is shown to be the most stable one from both the electronic and the lattice dynamical arguments, in agreement with the XRD and NPD-2K results. 4. Conclusions The combined high resolution synchrotron and neutron diffraction studies of MgNi2H(D)3 showed that during the hydrogenation of the hexagonal Laves type intermetallic compound MgNi2, it transforms into the orthorhombic MoSi2-type structure. Together with the ab initio calculation of the electronic structure, these studies revealed the mechanism of the metal-hydrogen bonding in the Cmca phase of MgNi2H(D)3.

Fig. 10. Heat capacities of MgNi2 and MgNi2H3 at constant volume (harmonic approximation: HA) and at constant pressure (quasi-harmonic approximation: QHA). Experimental results of Cp measured for MgNi2 are taken from the literature [20,21].

Fig. 11. Unit cell volume (solid curve, left vertical scale) and thermal expansion coefficient (dashed curve, right vertical scale) as a function of temperature calculated for MgNi2 and MgNi2H3 in the QHA approximation.


422

V.A. Yartys et al. / Acta Materialia 98 (2015) 416-422 [2] R. Kataoka, Y. Goto, A. Kamegawa, H. Takamura, Masuo Okada, High-pressure synthesis of novel hydride in Mg-Ni (-H) system, Mater. Trans. 47 (2006) 1957-1960. [3] R. Kataoka, A. Kamegawa, H. Takamura, M. Okada, High pressure synthesis of novel Mg(Ni1ÐxCux)2 hydrides (x = 0-0.2), Mater. Trans. 50 (2009) 1179-1182. [4] V.A. Yartys, V.V. Burnasheva, V.A. Yartys, N.V. Fadeeva, S.P. Solov'ev, K.N. Semenenko, Crystal chemistry of RT2H(D)x, RT5H(D)x and RT3H(D)x hydrides based on intermetallic compounds of CaCu5, MgCu2, MgZn2 and PuNi3 structure types, Int. J. Hydrogen Energy 7 (1982) 957-965. [5] V.E. Antonov, Phase transformations, crystal and magnetic structures of highpressure hydrides of d-metals, J. Alloys Compd. 330-332 (2002) 110-116. [6] L.G. Khvostantsev, V.N. Slesarev, V.V. Brazhkin, Toroid type high-pressure device: history and prospects, High Pressure Res. 24 (2004) 371-383. [7] H. Brian Toby, EXPGUI, a graphical user interface for GSAS, J. Appl. Crystallogr. 34 (2001) 210-213. [8] P.E. Bl?chl, Projector augmented-wave method, Phys. Rev. B 50 (1994) 17953- 17979. [9] G. Kresse, J. Furthm?ller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set, Phys. Rev. B 54 (1996) 11169- 11186. [10] G. Kresse, D. Joubert, From ultrasoft pseudopotentials to the projector augmented-wave method, Phys. Rev. B 59 (1999) 1758-1775. [11] J.P. Perdew, K. Burke, M. Ernzerhof, Emission in symmetric heavy ion reactions at subthreshold energies, Phys. Rev. Lett. 78 (1997) 1396-1399. [12] L. Chaput, A. Togo, I. Tanaka, G. Hug, Phonon-phonon interactions in transition metals, Phys. Rev. B 84 (2011) 094302. [13] K. Parlinski, Z.Q. Li, Y. Kawazoe, First-principles determination of the soft mode in cubic ZrO2, Phys. Rev. Lett. 78 (1997) 4063-4066. [14] A. Togo, F. Oba, I. Tanaka, First-principles calculations of the ferroelastic transition between rutile-type and CaCl2-type SiO2 at high pressures, Phys. Rev. B 78 (2008) 134106. [15] R.F.W. Bader, N. York (Eds.), Atoms in molecules: a quantum theory, Oxford University Press, Oxford, 1990. [16] G. Henkelman, A. Arnaldsson, H. J?nsson, A fast and robust algorithm for Bader decomposition of charge density, Comput. Mater. Sci. 36 (2006) 354-360. [17] R.V. Denys, V.A. Yartys, Colin J. Webb, Hydrogen in La2MgNi9D13: the role of magnesium, Inorg. Chem. 51 (2012) 4231-4238. [18] J.F. Herbst, L.G. Hector Jr., Structural discrimination via density functional theory and lattice dynamics: monoclinic Mg2NiH4, Phys. Rev. B 79 (2009) 155113. [19] B. Dorner, I.T. Belash, E.L. Bokhenkov, E.G. Ponyatovsky, V.E. Antonov, L.N. Pronina, Inelastic incoherent neutron scattering spectra from fcc NiH1.05, hcp CrH1.0 and hcp MoH1.2 at 15 K, Solid State Commun. 69 (1989) 121-124. [20] J. Wollam, E. Wallace, Magnetic susceptibility, heat capacity and third-law entropy of MgNi2, J. Phys. Chem. Solids 13 (1960) 212-220. [21] H. Feufel, F. Sommer, Thermodynamic investigations of binary liquid and solid Cu-Mg and Mg-Ni alloys and ternary liquid Cu-Mg-Ni alloys, J. Alloys Compd. 224 (1995) 42-54. [22] V.A. Yartys, O. Isnard, A.B. Riabov, L.G. Akselrud, Unusual effects of hydrogenation: anomalous expansion and volume contractions, J. Alloys Compd. 356-357 (2003) 109-113. [23] R.V. Denys, V.A. Yartys, Masashi Sato, A.B. Riabov, R.G. Delaplane, Crystal chemistry and thermodynamic properties of anisotropic Ce2Ni7H4.7 hydride, J. Solid State Chem. 180 (2007) 2566-2576. [24] V.A. Yartys, P. Vajeeston, A.B. Riabov, P. Ravindran, R.V. Denys, J.P. Maehlen, R.G. Delaplane, H. Fjellvõg, Crystal chemistry and metal-hydrogen bonding in anisotropic and interstitial hydrides of intermetallics of rare earth (R) and transition metals (T), RT3 and R2T7, Z. Kristallogr. 223 (2008) 674-689.

The diffraction data collected at 293 and 2 K are perfectly described by the Cmca unit cell, in which H atoms fill two types of sites, inside the deformed octahedral Mg4Ni2 site and in the infinite bended spirals -Ni-H-Ni-H- penetrating through the structure. The DFT calculations showed that the Cmca phase of MgNi2H3 is the most stable one among all considered alternatives, both from the electronic structure and the lattice dynamics arguments. Quasi-harmonic phonon calculations allowed estimating the heat capacity of the tri-hydride MgNi2H3. The Bader charge analysis for MgNi2H3 indicates an electronic transfer from Mg (Ð1.59eÐ) to Ni (+0.21eÐ), H1 (+0.55eÐ) and H2 (+0.31eÐ). The phonon dispersion curves of MgNi2H3 all show positive frequencies, indicating that the structure is mechanically stable. The calculated gross heat of formation of the Cmca-type MgNi2H3 is Ð37.3 kJ/mol-H2, which makes it by 3 kJ/mol-H2 more stable than the prototype structures tested earlier (Yartys et al. [1]). The stability of the crystal structure of MgNi2H3 is enhanced by the formation of the directional Ni-H covalent bonds supplemented by the electron transfer from Mg to both Ni and H atoms. Similar directional covalent Ni-H bonding has been observed earlier in the anisotropic CeNi3H2.7 [22] and Ce2Ni7H4.7 [23] hydrides and it also led to a significant rebuilding of the metal sublattices [24]. Thus, a strong Ni-H bonding interaction appears to be a key feature causing the rebuilding of the metal lattice during the process of formation of ternary hydrides. Acknowledgements This work is partly based on experiments performed at the Swiss spallation neutron source SINQ, Paul Scherrer Institute, Villigen, Switzerland. The work was supported by ERA Net Russia FP7 program (Project NOVELMAG # 225) and a Grant by the Program ``Elementary Particle Physics, Fundamental Nuclear Physics and Nuclear Technologies'' of the Russian Academy of Sciences. DFT and phonon calculations were performed using HPC resources from GENCI - CINES/IDRIS (Grants 2015-096175 and 2015-090189). References
[1] V.A. Yartys, V.E. Antonov, A.I. Beskrovnyy, J.-C. Crivello, R.V. Denys, V.K. Fedotov, M. Gupta, V.I. Kulakov, M.A. Kuzovnikov, M. Latroche, Yu.G. Morozov, S.G. Sheverev, B.P. Tarasov, Hydrogen-assisted phase transition in a trihydride MgNi2H3 synthesized at high H2 pressures: Thermodynamics, crystallographic and electronic structures, Acta Mater. 82 (2015) 316-327.