Atoms and Nanoparticles of TransitionMetals as Catalysts for Hydrogen Desorption fromMagnesiumHydride

The hydrogen desorption kinetics of composite materials made of magnesium hydride with transition metal additives (TM: Nb, Fe, and Zr) was studied by several experimental techniques showing that (i) a few TM at.% concentrations catalyse the H2 desorption process, (ii) the H2 desorption kinetics results stabilized after a few H2 sorption cycles when TM atoms aggregate by forming nanoclusters; (iii) the catalytic process occurs also at TM concentration as low as 0.06 at.% when TM atoms clustering is negligible, and (iv) mixed Fe and Zr additives produce faster H2 desorption kinetics than single additive. The improved H2 desorption kinetics of the composite materials can be explained by assuming that the interfaces between the MgH2 matrix and the TM nanoclusters act as heterogeneous sites for the nucleation of the Mg phase in the MgH2 matrix and promote the formation of fast diffusion channels for H migrating atoms.


Introduction
Because of the high hydrogen storage capacity of MgH 2 , ∼7.6 wt.%, low cost, and weight, magnesium is still currently investigated for H 2 storage applications.Most research efforts are dedicated to the improvement of the kinetic H 2 sorption process that is too slow even at temperatures larger than 500 K.The hydrogen desorption process results from many single steps taking place in series: the nucleation of the Mg phase in the MgH 2 parent phase and the diffusion of hydrogen atoms through transformed Mg and/or not transformed MgH 2 layers to the sample surface where the H 2 recombitative desorption occurs.The velocity of these reaction steps could be limited by the surface Mg oxide layer impeding the H 2 surface processes [1,2] or by the presence of a continuous MgH 2 layer acting as hydrogen diffusion barrier [3].On the contrary, the hydrogen absorption and desorption kinetic in Mg can be improved by microstructural refinements such as the reduction of the Mg or MgH 2 grain size: grain boundaries are, in fact, active nucleation sites for the formation or dissociation of the hydride phase and act as preferential diffusion paths for H atoms [4].
The major improvements are generally obtained by the use of a proper metal or metal oxide additive.Transition metals or transition metal oxides additives favour the H 2 dissociation [5] and can further improve the H 2 sorption kinetics in different ways.Nanocomposite materials formed by milling MgH 2 and Nb or Nb oxide powders are the most studied systems.Time-resolved X-rays synchrotron analysis suggested the transformation during H 2 desorption at 573 K of the NbH nanocatalyst in the metastable, H vacancy rich NbH 0.6 phases: the catalytic effect was thus explained as an enhanced transfer of H atoms from the MgH 2 matrix to the Mg surface by H diffusion through the catalyst layers [6].Recent studies on the structural analysis of MgH 2 samples catalysed by Nb [7] or Nb oxides [8] suggest the formation of Mg-Nb perovskite phases by reaction of Nb 2 O 5 and Mg: the perovskite was suggested to act as preferential "pathways" for hydrogen diffusion.
We have recently studied the H 2 desorption process from MgH 2 samples where the metallic additives, single or mixed atomic species, were completely included inside the MgH 2 matrix [9][10][11][12][13][14][15].These samples allow to specifically investigate the role of the additives, in the bulk processes, leading to the decomposition of MgH 2 phase and to the H 2 desorption process.In this paper we present a comprehensive review of our experimental results.
In particular, in Section 2 we review the studies on the hydrogen desorption kinetics from pure and Nb-catalysed MgH 2 , and then we discuss the role of the Nb additive when it is distributed inside the MgH 2 matrix at concentration ranging from <0.1 to about 5 at.%.
In Section 3 we will present the results of the experimental analysis dedicated to the structural and chemical-physical characterization of the Nb additive when dispersed in the Mg or MgH 2 matrix.
In Section 2 we present a study on the comparison between the catalytic effect of different TM additives contained inside the MgH 2 matrix at concentration ∼5 at.%.
Finally, in Section 3, we will present results on a novel route to enhance the hydrogen desorption kinetics from MgH 2 by using mixed TM additive.

Hydrogen Desorption Properties of Nb-Doped MgH 2 Samples
Samples of pure Mg and Nb-doped Mg with different concentrations of Nb (see Table 1) were deposited by RF magnetron sputtering in form of films with thickness ∼10- 20 μm on 5 cm diameter polished graphite wafers.The Nb doping was obtained by putting a few numbers of small Nb fragments on the surface of Mg target in such a way that sputtering process involves both Mg and Nb, that is, a co sputtering procedure was utilized to prepare Mg-doped samples [9].To prevent the surface oxidation and stimulate the H 2 dissociation at the Mg surface, samples were coated with a 15 nm thick Pd capping layer without interrupting the vacuum conditions in the deposition chamber.Pd is known to split hydrogen, and splitting is an important step in the hydrogen sorption process though it might be not the ratelimiting step.Indeed, at temperatures larger than 250-300 K, Pd is transparent to hydrogen [16].
The Nb concentration was measured, after deposition, by Energy Dispersion Spectroscopy (EDS) while Secondary Ions Mass Spectroscopy (SIMS) was employed with samples having Nb concentration lower than 1 at.% and to verify the uniform distribution of the Nb catalyst into the Mg layers [12].After deposition, Mg samples peeled off from the substrate and the self-supporting samples were introduced in a Sievert-type apparatus to study the hydrogen absorption and desorption kinetics: details on the sample preparation, activation procedure, and experimental analysis are described elsewhere [9,17].
In Figure 1 we present the hydrogen desorption kinetics from MgH 2 at 623 K in samples having different Nb concentrations.The figure shows a strong increase of the reaction velocity for all the Nb-doped Mg: the reaction halftime τ 1/2 , that is the interval time needed to have half of MgH 2 transformed to h-Mg, decreases from 3250 ± 50 s for pure MgH 2 to 110 ± 10 s for the 2-to-5 at.%Nb-doped MgH 2 .Intermediate values of τ 1/2 are observed at lower Nb concentrations: 290 ± 10 s for 1 at.% and 2390 ± 30 s for 0.06 at.%.A corresponding reduction of the incubation time, that is the typical time needed for the nucleation of the Mg phase, can also be observed in Figure 1 (arrows indicate the reacted fraction firstly evidenced that roughly corresponds to the incubation time).The desorption processes of Figure 1 have been satisfactorily described in the framework of the Johnson-Mehl-Avrami theory for solid-state transformation [18].The relevant effective parameters of the model, say, the values of the reaction order, n, and activation energy, E a , are reported in   The value n = 4 obtained for pure MgH 2 indicates that the continuous formation of stable Mg nuclei in the MgH 2 parent phase is the rate-limiting step in the H 2 desorption kinetics.n = 1 for MgH 2 samples with Nb content of 2-5 at.% indicates an instantaneous nucleation of the Mg metallic phase in the MgH 2 followed by its diffusional growth [9].The evaluated value of the activation energy, 51 ± 5 kJ/mol H, suggests that the diffusional growth is limited by H atomic diffusion in Mg layers: this value is, in fact, close to that of the activation energy for H diffusion in bulk Mg, 40 kJ/mol H [19]. Structural analysis by X-Ray Diffraction (XRD) and Extended X-ray Absorption Fine Structure (EXAFS) [12] evidenced, as we will report below, the formation of Nb:H nanoclusters.We thus explained the kinetic results by assuming that (a) the interface between the Nb:H nanoclusters and the MgH 2 matrix offers sites where the instantaneous nucleation of the Mg phase occurs (heterogeneous nucleation) and (b) H atoms dissociated from the Mg hydride migrate to the sample surface (where desorption occurs) through domains of transformed Mg.Kinetic data presented in Figure 1 indicate that the same behaviour is shown by samples with Nb dopant concentration ≥2 at.%.This result is relevant for application because it shows that the optimum catalyst concentration can be lower than 5 at.%, the concentration usually reported for ball-milled Mg [20].
Note that only a few other catalysts were tested at so low concentration [21].
Relevant differences on hydrogen desorption kinetics are observed in samples with Nb concentration lower than 2-5 at.%: the kinetic curve for 1 at.%Nb concentration is reproduced with n = 1.5 and E a = 78 ± 5 kJ/mol H.The XRD spectrum of such sample (see [12]) shows that Nb atoms form Nb:(H) nanoclusters with size nearly equal to that observed in samples with larger Nb content.Because of the lower Nb concentration, there is thus a larger distance between the Nb:H nanoclusters and a lower specific MgH 2 -Nb:H interface area as compared to the 2-5 at.%Nb-doped samples.The larger value of the activation energy (78 versus 51 kJ/mol) cannot be simply explained as an effective value describing the H 2 effusion kinetics of a material that consists of a fraction of Nb-catalysed MgH 2 and a fraction of pure MgH 2 : effusion simulation carried out by assuming that the hydride consists of a larger fraction of MgH 2 with embedded Nb clusters, and a lower fraction of pure MgH 2 cannot reproduce the observed H 2 desorption kinetics.On the contrary, by assuming that the observed kinetics still obeys a diffusion-controlled kinetics, the evaluated values of n = 1.5 and E a = 78 kJ/mol can be explained by observing that: (a) the presence of Nb:H clusters still provide a large density of Mg phase nucleation centers with low activation energy barriers (as in the case of 5 at.%Nb); this point is confirmed by the reduced values of the incubation time when compared with the pure Mg hydride; (b) the increased distance between the Nb clusters in the MgH 2 matrix limits the onset of interconnected Mg domains which favour the H atomic diffusion [9].
The evaluated value of the activation energy, 78 ± 5 kJ/mol H, is not much different from the value of the activation energy, 100 ± 10 kJ/mol H, for H diffusion in MgH 2 as reported by Fernández et al. [22].This suggests that the ratelimiting step in the H diffusion is given by the energy barrier to be overcome by the migrating species in not transformed MgH 2 phase.A final interesting result comes out from the analysis of the sample with the lower Nb concentration, 0.06 at.%.Here the clustering of the Nb atoms is a negligible process because the Nb concentration is close to the solubility limit of Nb in Mg, [Nb/Mg] ∼ 10 −4 [23].A catalytic effect of Nb on hydrogen desorption can be still observed as indicated by the reduction of the τ 1/2 parameter with respect to that observed in the undoped MgH 2 .After repeated H 2 absorption-desorption cycles, the catalytic properties decrease and the 0.06 at.%Nb-MgH 2 exhibits a value of the τ 1/2 parameter comparable to that of pure MgH 2 but lower incubation time.This means that the Mg nucleation process still results in acceleration.Because the relevant activation energy value of the desorption process is just the same as in the case of pure Mg (see Table 1), we may conclude that single Nb atoms may constitute seeds for heterogeneous Mg nucleation in the same way as homogeneous seeds operate in pure Mg.
In particular, we suggest that the atomic environment around the Nb impurity atom acts as nucleation seed for the Mg phase given the presence of local elastic strains due to the atomic size difference between solute (Nb) and solvent atoms (Mg, H).The decrease of the phase transition velocity after the first absorption-desorption cycles is explained by impurity segregation at extended defects, such as grain boundaries, that reduce the availability of heterogeneous nucleation seeds.
In conclusion, several different mechanisms influence the hydrogen kinetics in MgH 2 depending on the Nb atomic concentration: specifically, at concentration larger than 1 at.%,Nb forms nanoclusters that affect both the Mg nucleation process and hydrogen diffusion.At very low concentration, 0.06 at.%., Nb remains atomically dispersed (clustering of the Nb atoms should be negligible because the Nb concentration, as indicated above, is close to the solubility limit in Mg) and the only effect of Nb is on favouring the Mgphase nucleation.

EXAFS and TEM Analysis of Nb-Doped MgH 2
In order to further clarify the role of the Nb doping element, we studied the evolution of the chemical-physical state of the Nb atoms dispersed in the Mg matrix upon hydrogen absorption and desorption cycles.The analyses were carried out by EXAFS, XRD, and Transmission Electron Microscopy (TEM) analysis on three representative Nbdoped Mg samples: (i) as-deposited sample, (ii) activated MgH 2 samples after partial MgH 2 to Mg phase transition (hereafter, hydrogenated sample), and (iii) activated sample after complete MgH 2 to Mg phase transition (hereafter, dehydrogenated sample).Details on the measurements and data analysis are reported in [11].
Results show that stable catalytic effects of Nb are connected with the formation of Nb nanoclusters dispersed in the Mg matrix while no evidence of Nb oxidation could be deduced from previous analysis.
In Figure 2 we present the EXAFS spectra of the samples together with the spectrum of crystalline Nb: we observe that the spectrum of the as-deposited sample is different from the spectra of the hydrogenated and dehydrogenated samples which are similar to that of crystalline Nb.
In Figure 3 we show the moduli of the Fourier transformed spectra compared with the corresponding best-fit curves of the first two interatomic correlation signals: the spectrum of the as-deposited sample differs from that of the metallic Nb; the other two are quite similar, except for the fact that for the hydrogenated sample the peaks located at about 2.7 and 3.1 Å are shifted towards higher R (the interatomic distance) values with respect to the Nb foil spectrum.The results of the EXAFS analysis are reported in Table 2.
In the as-deposited sample, the average local structure around Nb is composed mainly of Mg atoms, at a distance that is shorter with respect to the corresponding value of the crystalline Mg and appears in agreement with the first Nb-Mg coordination distance obtained by simulating an Nb substitutional impurity in an Mg cluster.Besides this main signal, the analysis evidenced also an Nb-Nb correlation, as shown in Figure 3(b), where in the backtransformed k-space the filtered data and fit are compared and the different contributions to the fit are reported.The Nb-Nb correlation indicates the presence of Nb aggregates: the fact that the coordination number is low (about 1) and the Nb-Nb distance is shorter than for the bulk Nb strongly suggests that the clusters are composed of few Nb atoms.In the hydrogenated sample, the Nb-Mg signal is below the detectable limit; two Nb-Nb correlations are evident indicating the presence of Nb-containing nanoclusters: the Nb-Nb distances are ∼4.5% longer than the corresponding values in crystalline Nb.We ascribe this fact to the presence of H atoms into the Nb clusters increasing the interatomic distance: the local structure around Nb is in agreement with both the αNb-H phase (with an H/Nb atomic ratio of 0.9) [24] and with the orthorhombic βNb-H phase (with an H/Nb atomic ratio higher than 0.8) [25].It is worth noting that the coordination number is lower than the corresponding value in the bulk Nb, due to the fraction of Nb atoms located on the cluster surface and, possibly, to Nb atoms dispersed in the Mg matrix.After the hydrogen desorption process, the Nb is still aggregated in nanoclusters: in this case, the Nb-Nb coordination distances are those of the corresponding bulk phase: this indicates the H desorption from the Nb clusters (see [6] to gain insight on formation, decomposition, and general features on stability of NbH).The fact that the coordination numbers in this case are even lower than those of the hydrogenated sample could likely indicate a smaller size of the nanoclusters and/or a larger fraction of Nb atoms dispersed into the Mg matrix.As in the hydrogenated sample, the Nb-Mg signal is below the detectable limit for the EXAFS spectroscopy.
The XRD spectrum of the as-deposited sample [9] shows a preferential growth of the Mg film with the (0001) planes parallel to the substrate.After hydrogenation, the sample shows a complex pattern formed of many peaks; see Figure 4: the peaks can be attributed to a tetragonal MgH 2 phase with lattice parameters a = 0.4517 nm and c = 0.30205 nm (ICDD card n. 12-697) and to an orthorhombic β-NbH 0.89 phase with lattice parameters a = 0.484 nm, b = 0.490 nm, and c = 0.345 nm (ICDD card n. 7-263).The unindexed peaks are due to Mg layers that have lost the absorbed hydrogen.The mean grains size of the β-NbH 0.89 phase, estimated by using the Scherrer equation after deconvolution of the Mg(101) and NbH 0.89 (200) peaks, see the inset of Figure 4, is ∼20 nm.The presence of metallic Nb was never detected in the hydrogenated sample.The dehydrogenated sample shows a very simple XRD pattern where, apart from the unindexed Mg peaks, only the first three peaks of metallic Nb (bcc, a = 0.33066 nm, ICDD card n. 35-789) are present; see Figure 5.The mean grains size of the metallic Nb, calculated from the measurement of the FWHM of the Nb(110) peak, is ∼15 nm.It is worth to note that in this case the Mg-and Nb-hydrides are completely disappeared.
TEM observations performed on the as-deposited sample evidenced the preferential growth of the film, in agreement with the XRD results, as well as the absence of any nanocluster or precipitate inside the Mg grains.The size of these grains ranges from 0.1 to 1.5 μm.The electron diffraction pattern has a ring structure attributable to Mg.In this case, the few-atom aggregates detected by the EXAFS spectroscopy are most likely too small to give a significant contribution to the whole diffraction pattern.A completely different appearance is shown by the hydrogenated sample: spherical nanoclusters with dimensions ranging from 5 to 25 nm, not uniformly dispersed inside the grains, are clearly visible after few minutes of permanence under the electron  beam (a longer permanence induces the decomposition of the Mg-and Nb-hydrides).The electron diffraction pattern of the sample only shows the presence of Mg rings and a very weak ring attributable to metallic Nb.Finally, the dehydrogenated sample is stable under the electron beam, and its structure is formed of spherical nanoclusters with dimensions in the range 10-20 nm.These nanoclusters are not uniformly distributed inside the grains; see Figure 6, where the Nb particles appear as small black dots.The electron diffraction pattern of this sample is formed of spotty rings due to Mg and few well evident extra spots due to metallic Nb.
The present structural analysis thus shows that the accelerated H 2 desorption kinetics is connected to the presence of the Nb nanoclusters [12].As suggested by the analysis of the desorption curves, the role of the nanoclusters is connected to the accelerated nucleation of the Mg phase in the MgH 2 matrix: the interface between the Nb:H nanoclusters and the MgH 2 matrix favours the instantaneous nucleation of the Mg phase (heterogeneous nucleation).An embryo of the Mg phase can reduce its energetic formation cost at the MgH 2 -Nb:H interface as discussed in general framework; see [26,27].In addition, the MgH 2 -Nb:H interface can favour the Mg phase growth.Indeed, in diffusional phase transitions where transport of the atomic components occurs over paths longer than the interatomic distance, new phases are often observed to nucleate at grain boundaries of the parent phase.These extended defects result in fast diffusion channels for migrating atoms thus giving rise, in the lattice layers around them, to the critical solute concentration for the phase transition [26,27].In the present Nb-doped MgH 2 samples, the MgH 2 -Nb:H interface can play a similar role allowing an effective removal of the H atoms dissociated from adjacent MgH 2 layers and thus favouring the formation of Mg nuclei having critical size.

Comparison between Desorption Kinetics of Nb-, Zr-, and Fe-Doped MgH 2
In this section we present a comparative study on the H 2 desorption from the MgH 2 matrix doped with different metallic elements dispersed in atomic concentration, say, 2-5 at.% [13].Results on Nb were reported above and now we consider, for comparison, Fe and Zr because (i) they are transition metals (TM) with catalytic properties for the H 2 sorption kinetics in Mg and (ii) they do not form binary phases with Mg and show negligible solubility [23].Nb, Zr, and Fe have different "affinity" with hydrogen in the temperature range (400 K < T < 700 K).Nb forms interstitial alloys with hydrogen: the α phase is a random interstitial alloy of H in bcc Nb with H/Nb ratio <0.31.Zr forms stable phase: the δ phase having fcc structure with 1.31 < H/Nb < 2. Fe does not form any binary hydride phase and shows negligible H solubility [28].In these new composite samples, the atomic % concentration, as detected by EDS, was ∼5 at.% for Fe (the same as for Nb) and ∼3 at.% for Zr: these concentrations are larger than the maximum solubility of the TM in Mg, 0.00043 at.% for Fe at the eutectoid temperature of 920 K, close to 1.3 × 10 −4 for Nb in liquid Mg at 1500 K, and lower than 0.1 at.% for Zr at temperatures lower than 700 K [23].These equilibrium properties indicate that precipitation of the metal additive atoms is expected upon the thermal treatment leading to the sample activation at 623 K and/or during H 2 cycling.
The room temperature XRD spectra of the TM-doped samples, partially dehydrogenated, confirm cluster formation.While Fe remains in the metallic state, see Figure 7, Nb and Zr form hydride phase as shown by the NbH 0.89 and ZrH 1.66 reflection peaks in Figures 8 and 9, respectively: no evidence of metallic Nb or Zr was found in the partially desorbed samples.The obtained TM-H phases are appropriate to equilibrium phase diagram at room temperature [28].The Debye-Scherrer analysis of the XRD peak indicates the formation of TM nanoclusters having dimensions in the 10 ÷ 20 nm range while Mg and MgH 2 maintain the microcrystalline structure [9][10][11].The aggregation of TM atoms to form nanoclusters during cycling in hydrogen is confirmed by the evolution of XRD spectra from asdeposited to fully cycled samples [9,11] and by also comparing with EXAFS analysis [11].
In Figure 10 we present the isothermal H 2 desorption curves of the pure and metal-doped MgH 2 samples at the representative temperature of 623 K.The curves show a strong improvement of the H 2 desorption kinetics in the examined TM-doped samples when compared to the pure MgH 2 : at this temperature the interval time required for 50% transformation decreases from ∼3000 s for pure MgH 2 , to ∼200 s for the Zr-doped sample, and to ∼100 s for Nb-and Fe-doped samples.
In Figure 11 we present the TDS spectra of the pure and TM-doped MgD 2 samples after their complete activation.(For the TDS analysis deuterium was chosen rather than hydrogen to improve the signal-to-noise ratio: at 2% of the TDS peak height the signal-to-noise ratio of the m/e = 4 signal was better than 10 2 ).The most important information provided by the spectra is the position of the peak temperature T P at which the desorption rate reaches its maximum: this temperature allows a direct comparison of the influence of the TM additive on the thermal stability of the MgD 2 phase.The obtained spectra confirm the trend observed in the isothermal curves evidencing lower thermal stability of the TM-doped samples as compared to pure MgD 2 .The Fe-and Nb-doped sample show TDS peaks at ∼470 and ∼480 K, respectively while the Zr-doped samples shows TDS peak at ∼535 K. Pure MgD 2 shows the TDS peak at a larger temperature, close to 630 K.
The present H 2 (or D 2 ) desorption results show that the dispersion of different TM nanoclusters inside the MgH 2 matrix accelerates the H 2 desorption process and that TM nanoclusters distributed in the MgH 2 matrix catalyse the Mg phase nucleation even if they present different chemical affinity with hydrogen.This point is also confirmed by the strong reduction of the nucleation time (Figure 10) shown by the isothermal desorption curves of the TM doped samples (∼10 s for Nb and Fe and ∼50 s for Zr) as compared to that pertinent to the pure MgD 2 , ∼1000 s.
In conclusion, composite materials consisting on TMdoped Mg (TM: Nb, Fe, and Zr) with TM content of a few at.% concentration show accelerated H 2 desorption kinetics when compared to the pure MgH 2 samples (the diffusion of migrating hydrogen atoms has been measured by "in situ" EXFAS for the MgH 2 /Nb 2 O 5 system [29]); while in the pure MgH 2 the H 2 desorption kinetics is controlled by the nucleation and growth of the Mg phase, in TM-doped samples the H 2 release is controlled by the H diffusion through Mg layers.The improved H 2 desorption kinetics can be explained by the presence of extended interfaces between MgH 2 and TM nanoclusters acting as heterogeneous sites for the nucleation of the Mg phase and promoting the formation of fast diffusion channels for H migrating atoms.

Mixed Zr-Fe Catalysts
In the previous sections we observed that the presence of TM nanoclusters dispersed in the MgH 2 matrix accelerates the H desorption kinetics.Two additional effects were also evidenced by the experiments [12].First, in Mg samples with Nb content in the at.%level, the hydrogen desorption kinetics, when Nb does not form precipitates, was faster than that observed in the activated sample where Nb forms nanoclusters.Second, catalytic effects were also observed when the metallic additive was present at levels lower than 0.1 at.% in the Mg matrix, that is, close to the solubility limit of Nb in Mg.The desorption curves presented the same kinetics (nucleation and growth mechanisms) and activation energy as that of the pure MgH 2 samples but a shorter nucleation time: we thus suggested that also Nb atoms (or few-atom aggregates) dispersed in the Mg matrix act as nucleation centres for the Mg phase in the parent MgH 2 phase [12].These results suggest that a route to accelerate the H 2 desorption kinetics is to limit the aggregation of the additives in form of large clusters and to favour a better distribution of the catalyst in form of small clusters or few atoms aggregates into the MgH 2 matrix.To this purpose we have prepared MgH 2 samples with mixed TM additives: in this section we report on the hydrogen desorption from magnesium hydride samples containing Fe and Zr, at a few at.% concentration [14].
We have used Zr and Fe because they have catalytic effects when dispersed in form of nanocluster in the MgH 2 matrix and because they do not form binary phases with Mg matrix.Finally Zr is heavier than Fe and has larger atomic diameter, 4.32 versus 3.44 Å: they thus have different mobility values and we thus expect different clustering processes when sharing the same host matrix (Mg).
The Fe atoms will preferentially form clusters thus limiting the aggregation process of Zr atoms because Fe atoms can faster reach extended defects in the Mg (or MgH 2 ) matrix which are the typical nucleation centers.To experimentally study this process we prepared Mg samples with (i) Fe single additive (5 at.%), (ii) Zr single additive (3 at.%), and (iii) mixed additive with 7 at.%(∼5 and ∼2 at.% content for Fe and Zr, resp.).
XRD analysis of the as-deposited samples showed no evidence of metallic Fe and Zr: this means that the presence of the additive is in atomic form or in form of a few atoms aggregates as previously observed with Mg samples containing the Nb additive [11].Based on previous XRD analysis of TM-doped Mg, it is really difficult to suppose the presence of metallic amorphous phase.In Figure 12 we report the H 2 desorption data pertinent to fully activated MgH 2 samples with Fe-Zr mixed additive [18].For comparison in Figure 13 we present the H 2 desorption data at the representative temperature of 623 K pertinent to the MgH 2 samples with single catalyst: symbols ( ) for Zr, symbols ( ) for Fe.Symbols ( ) are pertinent to the sample with Fe-Zr mixed additives.
In Table 3 we report the kinetic parameters pertinent to samples containing the mixed additives and, as reference, those pertinent to samples with a single additive and to pure MgH 2 .We note that in presence of mixed additives the hydride to metal phase transition obeys a nucleationand-growth mechanism while in presence of a single catalyst (5 at.%Fe or 5 at.%Nb) the phase transformation has diffusive character as shown by the different values of the reaction order n and of activation energy E a .The most important result is the fact that the mixed additives induce faster desorption kinetic as compared to pure MgH 2 and to MgH 2 with Fe or Nb as single additive as shown by the τ 1/2 parameter.Table 3 also indicates that the hydride decomposition kinetic with mixed additives is very similar to that of the sample with 1 at.%Nb reported in Section 2: the order of reaction is the same, and the activation energy values are very similar.The much better catalytic effect of the mixed additives is related to the impressive increase of preexponential factor (A) growing from ∼10 4 to ∼10 8 sec −1 .We remark that the improved kinetics compared to that of the sample with single Fe or Zr additive cannot be attributed to the larger additive content because there is a saturation effect in the desorption catalysis with additive contents of ∼2 at.% [12].
In Figure 14 we present the XRD spectrum of a fully activated sample after H 2 desorption that provides information on the physical state of the Fe and Zr additives.In the spectrum we observe reflection peaks pertinent to Mg and the (110), (200), and (211) Bragg reflections of the α-Fe phase but no XRD peak attributable to Zr nor to any Fe-Zr phase.Because Zr and Fe as single additive cannot produce the fast desorption kinetics shown in Figure 12, we may suppose that the improved catalytic effect of the mixed additives has two causes: (i) aggregation of Fe into nanoclusters and (ii) atomic dispersion of Zr (that could also form aggregates but consisting on few atoms).We remark that the second effect has permanent character because, contrarily to the samples with single additive, no evidence of Zr clustering is observed also after repeated H 2 sorption cycles.By using the Williamson-Hall method [30] the size of the α-Fe crystallites was determined to be lower than 20 nm.We can also evaluate a low microstrain value of ε ∼ 6 × 10 −4 not observed in the Mg sample with single Fe additive.The intensity ratio of these diffraction peaks indicates that the α-Fe nanoclusters have a random distribution in the Mg matrix.The α-Fe clusters size is lower than in Mg samples with single Fe additive, ∼30 nm, as evaluated by the Bragg-Brentano analysis of the (110) reflection peak presented in the inset of Figure 14, line (a) (this is the only α-Fe reflection in the XRD spectrum).Both samples have the same Fe content, ∼5 at.%, and assuming that all Fe atoms form precipitates, we conclude that the sample with mixed additives presents a higher Fe cluster density.
In the Mg sample with Zr as single additive, after segregation upon diffusion, Zr atoms aggregate forming clusters [13].In the sample with mixed additives these Zr clusters are not observed: the larger Fe mobility suggests that a possible mechanism impeding the Zr aggregation is the Fe clustering in the disposable nucleation centres which anticipates the Zr precipitation.The lower size of the Fe clusters in the Mg sample with mixed additives can be explained by a larger density of available nucleation sites as compared to that in the Mg sample with single Fe additive: reasonably, the Zr few-atom aggregates may constitute precipitation centres for Fe atoms.To gain a deeper understanding on the kinetic processes in presence of mixed additives and to determine which is the optimum Fe and Zr atomic concentrations, we have prepared samples with fixed metal content, about 7 at.%,by changing the [Fe]/[Zr] atomic ratio [31].In Figure 15 we summarize the results by showing the size of the Fe nanoclusters, the Mg grain size, and the desorption rate constant k (T = 623 K) as function of the [Fe]/[Zr] ratio.From Figure 15 we observe that (i) the H 2 desorption rate constant attains its maximum value when the [Fe]/[Zr] atomic ratio is ∼1, (ii) Fe forms nanoclusters with 20 nm grain size independently on the [Fe]/[Zr] atomic ratio, (iii) at [Fe]/[Zr] ∼1.8 the Mg grain size attains its minimum value of about 100 nm, and (iv) there is not evidence of Zr atoms aggregates.
To explain the previous results, we suggest that the synergetic effect of the mixed Fe and Zr additives consists of (i) Fe aggregation forming nanosized clusters that act as Mg nucleation centers and promote the formation of interconnected Mg domains for fast H diffusion and (ii) Zr atoms playing most probably a role by favoring and stabilizing the aggregation of the Fe additive and as Mg grain refiner by limiting Mg grain growth [32]: grain refinement certainly contributes to increase kinetic favoring diffusion through grain boundaries.
In the future it would be interesting to investigate the possible existence of correlation effects on diffusion when different dopants are moving and clustering occurs as observed in [33].

Conclusions
We have studied the hydrogen desorption kinetics of TMdoped MgH 2 samples to analyse the catalytic role of TM atoms and nanoparticles dispersed into the magnesium hydride matrix.We observed that (i) TMs are able to enhance hydrogen desorption also at very low concentration (0.06 at.%) which means that they are atomically dispersed or form a few-atom aggregates, (ii) TM-doped samples show stable desorption properties after a number of H 2 sorption cycles when TM nanoclusters are formed with radius on the order of 10-20 nm, and (iii) mixed TM (in the present case Fe and Zr) can further enhance the desorption kinetics as compared to the single additive.
The improved H 2 desorption properties of the single TM additive are explained by the presence of interfaces between the MgH 2 and the TM nanoclusters acting as heterogeneous nucleation sites and promoting the formation of fast diffusion channels for H migrating atoms.In our case mixed TM (Fe-Zr) additive in the Mg matrix optimizes the distribution and size of TM nanoclusters contained inside the MgH 2 matrix.

Figure 2 :
Figure 2: Nb K-edge EXAFS spectra from Nb-doped (5 at.%)Mg samples; the spectra of the hydrogenated and dehydrogenated samples are pertinent to fully activated samples.

Figure 3 :
Figure 3: EXAFS analysis of the Nb-doped (5 at.%)Mg samples.(a) k 3 -weighted Fourier transform moduli (transformation range k = 3-13 Å−1 for the as-deposited sample, k = 3-16 Å−1 for the others) of the EXAFS spectra (markers) and corresponding best-fit curves (solid line) of the first two coordination signals (Nb-Mg and Nb-Nb for the as-deposited sample, two Nb-Nb for the others); no back-scatterer phaseshift correction was applied in the figure.(b) Fourier filtered spectrum (markers) in the fitting range and superimposed fit (solid line) of the as-deposited EXAFS spectrum; the residuum and the two single-scattering signals that contribute to the fit are reported.

Figure 4 :
Figure 4: XRD spectrum of the 5 at.%Nb-doped MgH 2 sample, partially dehydrogenated.The inset shows the deconvolution of the Mg (101) and NbH 0.89 (200) peaks: continuous line: experimental data, dashed line: fitting, and dotted line: envelope of the fitted curves.

Figure 11 :
Figure 11: TDS spectra of the pure and TM-doped MgD 2 samples after their fully activation.

Figure 12 :
Figure 12: Dehydrogenated fraction versus time at various temperatures for the Mg sample with mixed Fe-Zr additives at 623 K, O 613 K, 603 K, ♦ 593 K, 583 K, and 573 K.In the insert we present the Arrhenius plot of the desorption rate constant k: the straight line is the least square fit.

Figure 13 :
Figure 13: Dehydrogenated fraction versus time at 623 K of MgH 2 sample with Zr ( ) and Fe ( ) single additive and mixed Fe-Zr additives ( ).The lines are only guide for the eyes.

Figure 14 :
Figure 14: XRD spectra pertinent to Mg samples with mixed Fe-Zr additives (completely activated) after H 2 desorption: in the figure only the Fe reflection peaks are labelled and all others peaks are pertinent to h-Mg reflections.In the inset we present the Fe (110) reflection peak pertinent to a sample with only the Fe additive, line (a), and to a sample with mixed Fe-Zr additives, line (b).The thick line is the reference position of the Fe (110) peak [ICDD 06-0696].

Figure 15 :
Figure 15: Fe cluster size (triangles), Mg grain size (squares) and desorption rate constant (circles) at different [Fe]/[Zr] atomic ratios.The lines are only a guide for the eye.

Table
. The obtained different values suggest that different mechanisms control the H 2 desorption kinetics at different Nb concentrations.

Table 1 :
Evaluated values of the half-reaction time (τ 1/2 ) at T = 623 K, order of reaction (n), and activation energy (E a ) of MgH 2 samples with different Nb concentrations.

Table 2 :
Results of the Nb K-edge EXAFS analysis on Nb-doped Mg films; for each pair of atoms, N is the coordination number, R the interatomic distance, and σ 2 the Debye-Waller factor.The uncertainty in the fitting results corresponds to a confidence level of 68%.Crystallographic data for metallic Nb and the simulation of the local structure around an Nb substitutional impurity into the Mg lattice are also reported.

Table 3 :
Evaluated values of the half-reaction time (τ 1/2 ) at T = 623 K, order of reaction (n), activation energy (E a ), and pre-exponential factor (A) of pure and TM-doped (TM: Nb, Fe, and Zr + Fe) MgH 2 samples.