Phase Equilibria, Crystal Structure and Hydriding/Dehydriding Mechanism of Nd4Mg80Ni8 Compound

In order to find out the optimal composition of novel Nd-Mg-Ni alloys for hydrogen storage, the isothermal section of Nd-Mg-Ni system at 400 °C is established by examining the equilibrated alloys. A new ternary compound Nd4Mg80Ni8 is discovered in the Mg-rich corner. It has the crystal structure of space group I41/amd with lattice parameters of a = b = 11.2743(1) Å and c = 15.9170(2) Å, characterized by the synchrotron powder X-ray diffraction (SR-PXRD). High-resolution transmission electron microscopy (HR-TEM) is used to investigate the microstructure of Nd4Mg80Ni8 and its hydrogen-induced microstructure evolution. The hydrogenation leads to Nd4Mg80Ni8 decomposing into NdH2.61-MgH2-Mg2NiH0.3 nanocomposites, where the high density phase boundaries provide a great deal of hydrogen atoms diffusion channels and nucleation sites of hydrides, which greatly enhances the hydriding/dehydriding (H/D) properties. The Nd4Mg80Ni8 exhibits a good cycle ability. The kinetic mechanisms of H/D reactions are studied by Real Physical Picture (RPP) model. The rate controlling steps are diffusion for hydriding reaction in the temperature range of 100 ~ 350 °C and surface penetration for dehydriding reaction at 291 ~ 347 °C. In-situ SR-PXRD results reveal the phase transformations of Mg to MgH2 and Mg2Ni to Mg2NiH4 as functions of hydrogen pressure and hydriding time.

Scientific RepoRts | 5:15385 | DOi: 10.1038/srep15385 NdMg 8 Ni and Nd 5 Mg 41 phases. According to the experimental results, the determined phase diagram in the Mg-rich corner at 400 °C is plotted in Fig. 1(f).
The Nd-Mg-Ni ternary compounds are expected to be the hydrogen storage alloys with excellent properties because the uniform composition would leads to in-situ formation of the ultrafine MgH 2 -NdH x -Ni/ Mg 2 NiH 4 composites 11,21,26 . From the reported literature and our experimental work, it can be found that there are seven ternary compounds in the Nd-Mg-Ni system: Nd 4 Mg 80 Ni 8 , NdMg 8 Ni, NdMg 5 Ni 27 , NdMg 2 Ni 9 , Nd 2 MgNi 2 28 , NdMgNi 4 29 and NdMg 2 Ni 9 30 . The new compound Nd 4 Mg 80 Ni 8 is selected as the target alloy because of its highest content of Mg among those ternary compounds which indicates the maximum hydrogen capacity.
The crystal structure of Nd 4 Mg 80 Ni 8 . In order to determine the crystal structure of Nd 4 Mg 80 Ni 8 , the single phase was synthesized and examined by SR-PXRD. The actual composition of Sample #5 and #6 are Nd 4.08 Mg 89.98 Ni 5.94 and Nd 4. 45 Mg 84.64 Ni 10.91 , respectively. The SR-PXRD pattern of Sample #5 was indexed with a tetragonal unit cell using DICVOL06 31 . The structure solution started using the charge-flipping algorithm implemented in the program TOPAS v4.2 32 . The Ni and Nd atoms were easily located in the electron density maps. The structure was subsequently solved in the space group of I4 1 /amd (No. 141) by global optimization in direct space with 5 Mg atoms with no constraint using the program FOX 33 . Rietveld refinement was performed using TOPAS v4.2, and the refined lattice parameters were a = b = 11.2743(1) Å, c = 15.9170(2) Å, V = 2023.19(4) Å 3 . The diffraction profile fitted by Rietveld refinement using these parameters is shown in Fig. 2(a), with the agreement factors of R wp = 8.1%, R B = 6.4%, and GoF = 1. 44. The fitting result suggests that there were 87.8 wt.% Nd 4 Mg 80 Ni 8 and 12.2 wt.% Mg in the Sample #5. The details of the structure determination and crystallographic data are presented in Tables 1 and 2. The crystal structure of Nd 4 Mg 80 Ni 8 is shown in Fig. 2(b).
Nd 4 Mg 80 Ni 8 has a distinguished structure from other reported M-Mg-Ni (M = metal) ternary alloys. There is one symmetry independent Ni atom in the unit cell coordinates with six Mg atoms forming      Based on the TEM results, the mechanism of hydrogen-induced microstructure evolution can be revealed. When the alloy reacts with hydrogen, the Nd atoms firstly disassociate from Nd 4   absence of Nd atoms leads to the polyhedra of [NdMg 16 ] crumbling. The released Mg atoms make the structure highly disorder. According to the equilibrated phase diagram, the rest composition will shift to Mg + Mg 2 Ni two-phase region. The Mg and Ni atoms diffuse fast owing to the disordered structure. Then large particles of Mg and Mg 2 Ni (58 ~2 50 nm) form to reduce the Gibbs free energy of system. After the Nd 4 Mg 80 Ni 8 transforming to NdH 2.61 -Mg-Mg 2 Ni nanocomposites, the Mg reacts with hydrogen to generate MgH 2 and the hydrogen atoms dissolve in Mg 2 Ni to generate Mg 2 NiH 0.3 . The high density NdH 2.61 nanoparticles, numerous interfaces between MgH 2 and Mg 2 NiH 0.3 , and a large number of grain boundaries in the nanocomposites of NdH 2.61 -Mg-Mg 2 Ni may provide a great deal of hydrogen atoms diffusion channels and nucleation sites of hydrides. Thus, the NdH 2.61 -MgH 2 -Mg 2 NiH 0.3 nanocomposites should exhibit excellent H/D kinetics.

The thermodynamic and kinetic properties of H/D reactions in Nd 4 Mg 80 Ni 8 . The Sample #6
shows a good activation behavior at 350 °C. At the second H/D cycle, it reaches a maximum hydrogen capacity of 5.15 wt.% which is near to the theoretical value 5.18 wt.% H 2 . All the PCT curves at different temperatures shown in Fig. 6(a) manifests two flat plateaus, indicating that there are two phases reacting with hydrogen during the H/D processes. One of phases exhibits larger storage capacity and wider plateau of H/D reactions marked as the first plateau in Fig. 6(a). The second phase shows higher equilibrium pressure of hydrogen and narrower plateau marked as the second plateau. Table 3 gives plateau pressures, maximum hydrogen capacities at different temperatures and thermodynamic data for the different phases. The hysteresis factor defined as Hf = ln(P ab /P de    It absorbs 85% of the maximum hydrogen content above 250 °C within 5.8 min. After that the hydrogen absorption content increases slowly with prolonging time. At 1 h, the alloy absorbs 4.82 wt.% hydrogen at 350 °C, which is 93% of the theoretical hydrogen storage content. The sample exhibits good desorption kinetics as shown in Fig. 6(c). It releases the absorbed hydrogen thoroughly within 8.3 min when the sample is heated up to 291 °C.
Lots of scholars developed kinetic models for the gas-solid reaction, such as Jander model 44 , Ginstling-Brounshtein equation 45 , etc. Evard et al. 46 developed a mathematical model to describe the non-isothermal decomposition process of MgH 2 , which took into account relative rates of hydrogen desorption, chemical transformation on the MgH 2 -Mg interface and size distribution of the powder particles. In our previous work, Chou et al. [47][48][49] proposed a series of formulae concerning the isothermal kinetics of gas-solid reaction based on a real physical picture. All parameters in RPP model have clear physical meanings and the effects of temperature, pressure, particle size, sample shape, density change of resultant on the reaction fraction can be analyzed quantitatively. The treatment of this model avoids the multistep calculation error at multi-temperatures and multi-pressures 50 . up to now, the RPP model has   been successfully used in analyzing the H/D kinetics of Mg-Ni alloy 50-52 , LaNi 5 -based alloy 9 , Mg-LaNi 5 53 , La 2 Mg 17 -based composites 54 , etc. Therefore, the isothermal H/D kinetics of the Nd 4 Mg 80 Ni 8 are analyzed by fitting the observed curves using the RPP model. It is found that the rate controlling step is the diffusion of hydrogen in the hydride during hydrogenation by fitting the experimental data with Eq. (5).
ξ the reacted fraction equaling to the ratio of hydrogen absorption weight Δ m at time t to the maximum hydrogen absorption weight Δ m max , t c(d) the characteristic reaction time representing the required time of a completely hydriding or dehydriding of the sample, P H2 the partial pressure of hydrogen in gas phase, P eq the hydrogen partial pressure in equilibrium with hydride, and Δ E the activation energy. The characteristic time t c is regarded as a criterion for reaction rate: the larger the characteristic time, the slower the reaction rate. The corresponding squared correlation coefficient, r 2 , reflects the level of agreement between fitting curve and experimental data. Using Eq. (1) to fit the hydrogenation data, the calculated t c(d) decreases from 153.5 to 1.6 min when temperature increases from 100 to 300 °C, indicating that the hydriding reaction rate increases with the temperature rising. The apparent activation energy for hydrogenation is determined to be 82.3 kJ/mol by fitting the experimental data using Eq. (2). There is an interesting phenomenon that the fastest hydriding rate is observed at 300 °C for Nd 4 Mg 80 Ni 8 (t c(d) = 1.6 min) alloy within the investigated temperature range from 100 to 350 °C. It is known that both the forward reaction rate (hydriding reaction) and reverse reaction rate (dehydriding reaction) are accelerated with the increasing temperature. In addition, the hydrogenation of Mg and Mg 2 Ni is exothermic, while the dehydrogenation reaction is endothermic. The increase of temperature is propitious for the reverse reaction. If the reverse reaction rate increases more rapidly than the forward reaction rate, an apparent fastest hydriding rate would be found in the temperature range.
In order to compare the hydriding rate of Nd 4 Mg 80 Ni 8 with that of other Nd-Mg-Ni alloys in literatures 12,13,16,18,26 , the calculated results of characteristic reaction time are listed in Table S2.   Fig. 6(d) shows that the optimal Nd-Mg-Ni alloy is the Nd 4 Mg 80 Ni 8 designed in present work by considering the hydriding kinetics and hydrogen storage capacity.
The rate controlling steps are surface penetration (sp) of hydrogen atoms for dehydriding reaction at 291 ~ 347 °C through fitting the experimental data with Eq. (7). The calculated t c(sp) are 8.8, 3.0 and 1.5 min with temperature increasing from 291 to 347 °C. This means that the dehydriding reaction rate increases with the temperature rising. The activation energy for dehydrogenation is calculated to be 97.5 kJ/mol, which is much smaller than 160 kJ/mol for ball milled pure MgH 2 55 , 124.6 kJ/mol for induction melted Mg 90 Ce 5 Ni 5 alloy 8 , and comparable to 104 kJ/mol for the as-cast CeMg 3 21 . Combining Eqs (3 and 4), the dehydriding kinetic curves at any other temperatures can be predicted by RPP model as follows: where R the gas constant, T temperature in Kelvin, and 5.0 wt.% the largest desorption hydrogen content from experimental. The calculated and predicted curves are shown in Fig. 6(c), which suggests that the theoretical calculation agree well with experimental data. The cycle life kinetics was examined at 300 °C under 3.0 MPa H 2 . The hydriding behaviors of the 1 st ~ 5 th , 10 th , 39 th and 58 th cycles are showed in Fig. 7(a). It can be seen that the hydriding rate increases with the increase of cycle times from 1 st to 5 th . After the 5 th cycle, the hydriding rate becomes very fast. Figure 7(b) shows the hydrogen storage capacity versus cycle times. The hydrogen capacity increases sharply from 2.36 to 4.54 wt.% in the first 3 cycles and then increases gradually to the maximum value of 4.77 wt.%. Until the 58 th cycle, the hydrogen storage capacity still remains stable, which suggests that the Nd 4 Mg 80 Ni 8 has a good cycle ability. In order to investigate the relationship between grain size and cycle times, the XRD pattern was collected after the 1 st ~ 5 th and 10 th cycles, shown in Fig. 7(c). The samples were vacuumed at 300 °C for 2 h. A part of powders after the 5 th cycle were further vacuumed at 350 °C for 2 h. The calculated grain size versus cycle numbers is showed in Fig. 7(d). It can be seen that the grain size of NdH 2.61 increase slowly with the cycle number, but the grain sizes of Mg and Mg 2 Ni decrease in the first 3 cycles and then increase with the increase of cycle number. The Nd 4 Mg 80 Ni 8 disappears until the 4 th cycle and the phase fractions of Nd 4 Mg 80 Ni 8 in the first 3 cycles are 7.0 ± 0.6, 4.1 ± 0.2 and 2.5 ± 0.5 wt.%, respectively. Although most of Nd 4 Mg 80 Ni 8 decomposed in the 1 st cycle, the hydrogen absorption content is only 2.36wt.%. It suggests the generated Mg and Mg 2 Ni didn't absorb hydrogen fully. Therefore, in the 2 nd and 3 rd cycle, the uncompleted phase transformation of Mg ↔ MgH 2 and Mg 2 Ni ↔ Mg 2 NiH 4 reduced the grain size of Mg and Mg 2 Ni. After the sample is completely activated, the phase transformation can be finished at the initial stage of H/D process. The long holding time at this temperature leads to the growth of grain size. Therefore, after the 4 th cycle the grain sizes of Mg and Mg 2 Ni growth obviously.
The grain size of the sample further vacuumed at 350 °C is larger than that dehydriding at 300 °C. It suggests that the grain size grows with the raise of temperature and the extension of time. The growth of NdH 2.61 is slowly with cycle times, but the growth of Mg is obviously from 60 ± 2 nm after the 3 rd cycle to 87 ± 3 nm after the 10 th cycle. The grain size of Mg vacuumed at 350 °C is about 83 ± 3 nm which is smaller than the value of Mg (150 nm) reported by Denys et al. 20 at the same temperature. This is because it was pure Mg sample used in their study, while the well-distributed NdH 2.61 and Mg 2 Ni in present work can restrain the growth of Mg 21 .
The phase transformation of NdH 2.61 -Mg-Mg 2 Ni nanocomposites during hydrogenation. The phase evolution mechanism of RE-Mg-Ni alloys during hydrogenation/dehydrogenation process was well clarified by Denys et al. 20,24,25 combining in-situ SR-PXRD. The effect of solidification rate on the microstructure of alloy, phase structural and microstructural state of constituents during reversible process of synthesis and decomposition of hydrides, and kinetic mechanism during hydriding and dehydriding process were studied in detail. Inspiring by their work, the in-situ SR-PXRD assisted with Rietveld refinement was also applied to study the mechanism of phase transformation under different hydrogen pressures and at different time. The SR-PXRD patterns under different hydrogen pressures at 350 °C are shown in Fig. 8(a). The indexation of the pattern of the activated powders indicates the existence of NdH 2.61 , MgH 2 and Mg 2 Ni. Combined with the TEM results, the reaction taking place during the first hydrogenation is assumed to be:  The calculated fraction of each phase versus pressure is plotted in Fig. 8(b).  Fig.6 (a) . .
. Seen from the isothermal hydriding kinetic curves as shown in Fig. 6(b), the hydriding process of the alloy above 200 °C can be separated as two stages. The first was the rapid hydriding stage, while the second stage exhibited relatively slow hydriding rate. The phase composition during the hydriding process at 300 °C under 2.00 MPa H 2 was analyzed by in-situ SR-PXRD, shown in Fig. 8 The high density grain boundaries in the nanocomposites of NdH 2.61 -Mg-Mg 2 Ni provided a great deal of hydrogen atoms diffusion channels and nucleation sites of hydrides, which greatly enhances the H/D kinetics and improved the cycle ability. The grain size of NdH 2.61 grows slowly with cycle number, but the grain sizes of Mg and Mg 2 Ni decrease in the first 3 cycles, and then increase with the increase of cycle times. The growth of grain size is related with temperature and vacuum time. The kinetics mechanism is analyzed by RPP model, which suggests that the rate controlling step was diffusion for hydrogenation and surface penetration for dehydrogenation.

Experimental Methods
The preparation and examination of the equilibrated alloys. The Nd-Mg-Ni samples were prepared by a medium frequency induction furnace using blocks of Nd (≥ 99.99 wt.%), Mg (≥ 99.99 wt.%) and Ni (≥ 99.99 wt.%) as the starting materials. The as-cast samples were enclosed by tantalum foils for subsequently sealing in evacuated quartz tubes. The samples were annealed at 400 °C for 30 days and then quenched in ice-water. The sample compositions and heat treatment conditions were listed in Table S1.
The actual composition of each alloy was determined by inductively coupled plasma atomic emission spectrometry (ICP). The microstructure and composition of phases in the bulk samples were investigated by HITACHI SU-1500 scanning electron microscopy (SEM) equipped with energy dispersive X-ray Scientific RepoRts | 5:15385 | DOi: 10.1038/srep15385 spectrometer (EDS). The phase composition of annealed samples were characterized by X-ray diffraction (XRD) using 18KW D/MAX2500V + /PC diffractometer with Cu Kα radiation.
The solution of crystal structure. According to the average composition of at.% 4.86Nd-87.46Mg-7.68Ni detected from EDS, the Nd 4 Mg 80 Ni 8 compound (Sample #5) was prepared by annealing an induction melted ingot at 400 °C for 30 days followed by ice-water quenching. The actual composition determined by ICP located in the region of Nd 4 Mg 80 Ni 8 + Mg two-phase equilibrium. SR-PXRD data for Sample #5 were collected at a wavelength of 0.8262 Å by a Mythen-II detector on powder diffraction beamline, Australian synchrotron. The powdered samples were loaded into pre-dried 0.7 mm quartz capillaries fitted with a flow cell under an atmosphere of argon. The Rietveld refinement was performed using TOPAS v4.2 32 .
The microstructure of Nd 4 Mg 80 Ni 8 and its hydrogen-induced microstructure evolution. A bulk sample of Nd 4 Mg 80 Ni 8 compound (Sample #6, actual composition is Nd 4.5 Mg 84.6 Ni 10.9 ) with size of 3 × 3 × 2 mm was polished to obtain a smooth surface. The sample was consist of major phase Nd 4 Mg 80 Ni 8 and minor Mg 2 Ni. Then the bulk sample was incompletely hydrogenated at 350 °C under 2.0 MPa H 2 for 1 h. The microstructure of the sample was examined by FEI Helios Nanolab 600i dual beam focused ion beam FIB. In order to compare the microstructure and phase composition of Nd 4 Mg 80 Ni 8 and hydrogenated sample, two thin slices with thickness less than 100 nm were cut from the annealed Nd 4 Mg 80 Ni 8 and hydrogenated sample respectively by FEI Helios Nanolab 600i dual beam FIB. SAED patterns and HR-TEM images were collected by Tecnai G2 F20 S-Twin TEM.
The measurement of H/D properties. The H/D properties of annealed Nd 4 Mg 80 Ni 8 (Sample #6) was tested using automatic PCT characteristics measurement system from SUZUKI HOKAN. CO., LTD. in Japan. The Nd 4 Mg 80 Ni 8 was mechanically crushed into micro-particles (− 100 mesh, < 150 μ m) and activated at 350 °C under 4.0 MPa H 2 for hydrogen absorption and at the same temperature in vacuum for hydrogen desorption. The PCT curves were measured at 250 ~ 350 °C with the maximum equilibrated time of 40 min. The hydrogen absorption kinetics were examined at 100 ~ 350 °C under initial hydrogen pressure of 3.4 MPa. Before hydrogenation the sample was kept in vacuum at 350 °C for 2 h to ensure its complete dehydrogenation. The isothermal dehydriding kinetics was examined at 291 ~ 347 °C in vacuum after the sample completely hydriding at 350 °C for 2 h under initial hydrogen pressure of 3.4 MPa. The cycling behavior of the Nd 4 Mg 80 Ni 8 was determined at 300 °C. The time for hydriding under initial pressure of 3.0 MPa H 2 was 2 h and for dehydriding in vacuum was 2.8 h. In order to study the relationship between grain size and cycle number, the XRD pattern was collected after the 1 st ~ 5 th and 10 th cycles. The samples were vacuumed at 300 °C for 2 h and then air cooled to room temperature. A part of powders after the 5 th cycle were further vacuumed at 350 °C for 2 h to observe the growth of grain size. The XRD patterns were collected by Bruker AXS D8 diffractometer with Cu Kα radiation. The sizes of the crystallites in the samples were calculated from the refinements of XRD patterns using Scherrer equation.
The evolution of phase composition during hydrogenation. The in-situ SR-PXRD data of Nd 4.5 Mg 84.6 Ni 10.9 powders were collected at by a wavelength of 0.8262 Å by a Mythen-II detector on powder diffraction beamline, Australian synchrotron. The completely dehydrogenated Nd 4.5 Mg 84.6 Ni 10.9 powders were loaded into pre-dried 0.7 mm quartz capillaries fitted with a flow cell under an atmosphere of argon. The sample was heating to the set temperatures by a Cybostar hot air blower with heating rate of 20 °C/min under vacuum, then hydrogen was imported in and the detector started to collect the XRD data. Before heating, the quartz capillaries didn't scrubbing with argon, which led to some oxidation of the sample during heating process. Therefore, about 10.0 wt.% Nd 2 O 3 emerged at the initial stage of examination. Then the quantity of Nd 2 O 3 didn't change any more. The phase transformation of Mg and Mg 2 Ni during the hydriding process still can be observed. The diffraction data under different hydrogen pressures from 0.0 to 2.0 MPa were collected at 350 °C and proceeded for 4 min at one pressure. Before every collection, the sample was kept under this pressure for 30 min. The diffraction data at different time was measured every 2 min at 300 °C under a constant pressure of 2.0 MPa H 2 .