Available online at www.sciencedirect.com
Intermetallics 16 (2008) 179e187www.elsevier.com/locate/intermet
Structural, magnetic, electrochemical and hydrogen absorptionproperties of GdyMg2�yNi4�xAlx compounds with
0.4< y< 2 and 0< x< 1.2
Jean-Gabriel Roquefere a, Bernard Chevalier a, Rainer Pottgen b, Naoyoshi Terash*ta c,Kohta Asano d, Etsuo Akiba d, Jean-Louis Bobet a,*
a Institut de Chimie de la Matiere Condensee de Bordeaux (ICMCB), CNRS-UPR 9048, Universite Bordeaux1, 87 Av. Schweitzer, F-33608 Pessac, Franceb Institut fur Anorganische und Analytische Chemie Corrensstraße 30/36, D-48149 Munster, Germany
c Japan Metals & Chemicals Co., Ltd., 232 Oguni-machi, Nishiokitama-gun, Yamagata 999-1351, Japand National Institute of Advanced Industrial Science and Technology (AIST), AIST Central-5, 1-1-1 Higashi, Tsukuba, Ibaraki 305-8565, Japan
Received 5 June 2007; received in revised form 19 July 2007; accepted 5 September 2007
Available online 19 November 2007
Abstract
A solid solution with the C15b structure (MgCu4Sn-type) exists as a series from GdNi2 to Gd0.4Mg1.6Ni4 (Gd0.2Mg0.8Ni2). These compoundswere successfully elaborated by (i) Mechanical Alloying (MA) and (ii) melting followed by a subsequent annealing. It was also possible to syn-thesise the solid solution GdMgNi4�xAlx with x up to 1.2. The product is highly dependent on the elaboration route which induces a drasticchange in both chemical and physical properties (i.e. magnetism, electrochemistry, and structural disorder). A direct relationship between allthese property modifications was established. Finally, it is shown that different compounds have different behaviours towards hydrogen sorptionregarding both the hydrogen uptake and the thermodynamic results.� 2007 Elsevier Ltd. All rights reserved.
Keywords: A. Laves phases; B. Crystal chemistry of intermetallics; B. Hydrogen storage; C. Mechanical alloying and milling
1. Introduction
Hydrogen is an ideal non-polluting vector of energy for thefuture. However, its mass production and storage have to beimproved. To get rid of the heavy and bulky bottles, hydrogencould be stored in the solid state. Such a material must presenta high density of hydrogen and provide a low cost and easyavailability of raw materials. The best candidates for hydrogenstorage in the solid state are Metal Organic Frameworks [1],the carbonaceous systems [2], the chemical hydrides [3], mag-nesium [4], or the intermetallics [5], among others.
Metallic magnesium is cheap, abundant in the earth’s crustand can form the hydride MgH2 (i.e. high reversible capacity
* Corresponding author. Tel.: +33 54000 2653; fax: +33 54000 2761.
E-mail address: [emailprotected] (J.-L. Bobet).
0966-9795/$ - see front matter � 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.intermet.2007.09.002
of 7.6 wt.%). However, the hydriding reaction is very slow andsome catalysts have to be added onto the magnesium surfaceso as to enhance the reaction kinetics [6,7]. Intermetallicssuch as TiNi, ZrMn2 or LaNi5 [8] can absorb and desorb hy-drogen at room temperature under moderate pressures but theirweight capacities remain low (typically 1.5 wt.%). Regardingthe advantages and drawbacks of Mg on one hand and inter-metallics on the other hand, some authors have tried to studythe composites MgeABx [9] using an intermetallic as a cata-lyst, this latter being able to form a hydride around themagnesium grains.
The addition of catalysts is pertinent to modify the kineticsbut it doesn’t have any impact on the thermodynamics. Tochange the stability of Mg, it is necessary to modify the chem-ical bonds. For example, GdNi2 forms a rather stable hydride[10] whilst MgNi2 does not form a hydride at ambient condi-tions (25 �C and 0.1 MPa H2) [11] so that in a straightforward
mailto:[emailprotected]
180 J.-G. Roquefere et al. / Intermetallics 16 (2008) 179e187
approach, the substituted compounds GdyMg2�yNi4 shouldform reversible hydrides under ambient conditions. The slowhydriding reaction of MgNi2 when modified with the rapid re-action of GdNi2 could result in a material of technologicalimportance.
In a preliminary study on GdMgNi4 [12] it has been provedthat the substitution of Al for Ni leads to the compositionGdMgNi3Al, which remains in the space group F43m (C15bstructure, MgCu4Sn-type), and the hydrogen capacities werethen improved in comparison with the original compoundGdMgNi4. To our knowledge, some authors have also studiedsuch pseudo-AB2 Laves phases but have only focused on elec-tronic properties [13], crystal structure analysis [14], electro-chemical performances [15], or hydrogen absorption [16].
In this study, the compounds GdyMg2�yNi4 and GdMgNi4�xAlx were elaborated using both Mechanical Alloying (MA)and heat treatment methods. The influence of the synthesismodes on the physical and chemical properties was studied.We finally establish some relationships between synthesismethods and measured properties.
2. Experimental details
Pure elemental powders of Mg, Gd, Ni and Al (STREMChemicals, purity >99.5%) were mixed in an argon dry boxin the appropriate stoichiometry, put in a stainless steel con-tainer hermetically closed (weight ratio mpowders/mballs¼ 1:20)and ball-milled using a Fritsch P5 equipment with a plateaurotation speed of 250 rpm. The milling was performed for10 h continuously and the samples were then collected underargon, cold pressed into pellets of about 1 g and finally an-nealed at 650 �C for 1 h in an evacuated sealed silica tubeplaced in an electric furnace (experimental conditions wereoptimized in Ref. [12]).
Same compositions were melted (i) in tantalum cruciblesealed under argon and melted at 1050 �C followed by anneal-ing at 650 �C for 2 h or (ii) in alumina crucible using a vacuumhigh frequency induction furnace and annealed at 1323 �C for10 h under an argon atmospheric flow, as described inRef. [17].
The structural properties of the products were analyzed byX-ray powder diffraction using a Philips PW 1050 diffractom-eter with Cu Ka1 radiation (l¼ 0.15405 nm). In order to de-termine the intermetallic and the hydride structures, theleast-squares Rietveld refinement method was applied, thediffraction patterns being analyzed by a whole pattern fittingprocedure using the program FULLPROF [18].
The chemical composition and hom*ogeneity of the sampleswere checked by Electron Probe Micro-Analysis (EPMA) us-ing a CAMECA SX-100 instrument. The quantitative determi-nation was performed on the basis of intensity measurementsof Mg Ka1, Gd La, Ni Ka1 and Al Ka1 X-ray emission linesusing GdNi, MgNi2 and Al as reference compounds.
Hydrogen sorption properties were investigated with theuse of an automatic Sievert type volumetric apparatus in thetemperature range between 243 and 293 K and a pressurerange from 0.03 to 4 MPa [19].
Magnetizations were determined between 4.2 and 300 Kusing a SQUID magnetometer with fields from 0.1 T for mag-netization measurements to 3 T for susceptibility measure-ments in the paramagnetic range.
NieMH negative electrodes were prepared by mixingGdMgNi4 and graphite (mass ratio¼ 1:1) with 1 wt.% Teflonand by cold pressing this mixture at 100 MPa on a nickelfoam. The electrolyte was KOH (8 M) and the positive elec-trode was made of Ni(OH)2/NiOOH with a capacity 10 timeslarger than that of the negative electrode. Charging and dis-charging were limited, respectively, in time (6 h) and in poten-tial (0.9 V). On the basis that 4 H atoms were exchangedduring a cycle [15], a theoretical capacity of 257.3 mA h/gwas calculated and thus a constant current of 51.46 mA/gwas imposed (i.e. C/5).
3. Results and discussion
3.1. Synthesis and structural analysis
The compounds GdyMg2�yNi4 and GdMgNi4�xAlx wereprepared by both Mechanical Alloying (MA) and melting asdescribed previously. Because of the high saturating vapourpressure of Mg at high temperature, Mechanical Alloying fol-lowed by a heat treatment appears as an appropriate synthesismethod. As an example, Fig. 1 shows a picture of a pellet of‘‘GdMgNi4’’ with the composition of different noticed phasesas well as an XRD pattern of the same sample.
The sample appears to be very hom*ogeneous even if somesmall impurities (mainly GdNi5) were highlighted on EPMA(but not detectable on XRD pattern proving the presence ofvery small amount of impurity). The measured compositionis very close (almost in perfect agreement) to the nominalone with no Mg loss being detected. All of the diffractionpeaks are indexed in the F43m space group and the cell param-eter (i.e. a¼ 7.0350(7) A) is in good agreement with thepublished values [14].
It has been shown that the compound GdMgNi4�xAlx existsas a single phase up to x¼ 1.2. The purpose of replacing nickel(r¼ 1.24 A) by aluminium (r¼ 1.43 A) is for the expansion ofthe lattice cell as well as the sites available to host hydrogen. Asa result, a decrease of the equilibrium pressure is expected likethat in GdNi5�xAlx or LaNi5�xAlx [20]. The XRD patterns ofGdMgNi4, GdMgNi3.5Al0.5 and GdMgNi3Al are shown inFig. 2 and reveal a shift of the diffraction peaks towards thesmaller angles due to the presence of aluminium. Some smalladditional peaks also appear and they can be attributed toGdNi5 and MgNi2 impurities. The values of the cellparameters confirm the Vegard law that was established inour preliminary study (i.e. a¼ 7.038þ 0.102x) [12].
On a similar approach, the synthesis of GdyMg2�yNi4 waspossible from y¼ 0.4 to 2 by both melting and MA. The ele-ments Gd and Mg are strongly electropositive and have similarmetallic radii (i.e. r(Gd)¼ 1.78 A and r(Mg)¼ 1.60 A). How-ever, one is a light element and the other is a rare earth element,hence they exhibit very different electronic properties. Asa consequence no miscibility range is noticed in the GdeMg
Fig. 1. (a) EPMA analysis of ‘‘GdMgNi4’’ with the measured compositions and (b) XRD pattern of the same sample.
181J.-G. Roquefere et al. / Intermetallics 16 (2008) 179e187
binary diagram but just a couple of defined compounds are ob-served [21]. On the other hand, it has been noticed numeroustimes that a third element C could stabilize by electronicand/or steric effect an abnormal solubility between A and Bin a ternary ABC compound. For that reason, a solid solutionof LaxMg1�xNi2 is possible with 0.33< x< 1 [22] while nototal miscibility area exists in the binary diagram of LaeMg[21]. De Negri et al. confirmed such an existence by establish-ing the ternary diagram of LaeNieMg [23]. The XRD patternsof GdyMg2�yNi4 with different compositions are presented inFig. 3.
The X-ray diagram of Gd0.4Mg1.6Ni4 (Fig. 3a) shows thepresence of more than just the single phase of C15b struc-ture-type. It is possible to highlight the presence of MgNi2(space group P6/mmm). This statement suggests that y¼ 0.4is beyond the limit value of demixion. Fig. 3b correspondsto Gd0.6Mg1.4Ni4 prepared by Mechanical Alloying. Thebroadness of the diffraction peaks is due to the loss of crys-tallinity induced by ball-milling. Anyway the compound isalmost single-phased as confirmed by EPMA analysis
10 20 30 40 50 60 70 80
200
400
600
800
1000
1200
1400
1600
1800
In
ten
sity (a.u
,)
2θ (°)
(a)
(b)
(c)
440
Fig. 2. XRD patterns of (a) GdMgNi4, (b) GdMgNi3.5Al0.5 and (c)
GdMgNi3Al. Shift to smaller angles is highlighted for the (440) planes as
an example.
(measured composition: Gd0.60Mg1.40Ni3.98). Also, in othercompositions (Fig. 3c and d), the samples are almost sin-gle-phased. Fig. 3 as a whole allows us to notice a shift ofthe diffraction peaks towards the smaller angles when y in-creases. As stated previously for GdMgNi4�xAlx, an increaseof the cell parameter with the Gd content is noticed. The plotr (g/cm3) as a function of the compositions is established andpresented in Fig. 4.
The values related to GdyMg2�yNi4 are perfectly alignedwith those of GdMgNi4�xAlx proving the validity of the Ve-gard law. Beyond the observed linearity, one point is of a spe-cial interest: the point corresponding to the weakestvolumetric weight (i.e. z6.8 g/cm3) is related to both the sol-ubility limits. In the GdyMg2�yNi4 configuration, it meansy¼ 0.4 (Gd0.4Mg1.6Ni4) which is the value above whicha demixion is observed with notably the apparition ofMgNi2 as noticed previously with the DRX/EPMA analysis.In the configuration of GdMgNi4�xAlx the same point meansx¼ 1.2 (GdMgNi2.8Al1.2) which is exactly the solubility limitof Al in Ni 16e sites as seen previously [12].
Fig. 3. XRD patterns of GdyMg2�yNi4 with (a) y¼ 0.4, (b) y¼ 0.6, (c) y¼ 1.0
and (d) y¼ 1.4.
0,0 0,8 1,2 1,6 2,0 2,4
6,5
7,0
7,5
8,0
8,5
9,0
9,5
10,0
10,51,5 1,0 0,5 0,01,2
ρ (g
/cm
3)
y in GdyMg
(2-y)Ni
4
0,4
Solu
bilit
y lim
itsx in GdMgNi
(4-x)Al
x
Fig. 4. Calculated densities of GdyMg2�yNi4 and GdMgNi4�xAlx as a function
of both x (bold circles) and y (white squares).
20 25 30 35 40 45
2000
4000
6000
8000
10000
12000
(c)
(b)
222
311
220
200
In
ten
sity (a,u
,)
2θ (°)
2θ (°)
111
(a)
20 21 22 23 24 25 26 27 τ = 50%
200
111
τ = 0%
(d)
Fig. 5. Theoretical XRD patterns of GdMgNi4 with (a) Gd in 4a and Mg in 4c
sites (exchange rate t¼ 0%), (b) 0.75 Gdþ 0.25 Mg in 4a and
0.25 Gdþ 0.75 Mg in 4c sites (exchange rate t¼ 25%), (c) 0.5 Gdþ 0.5 Mg
182 J.-G. Roquefere et al. / Intermetallics 16 (2008) 179e187
It can be assumed that the area limited by the two dasheddot lines (yellow streaks) corresponds to a composition rangepotentially usable for hydrogen storage materials in practicalapplications (for interpretation of the references to colour ofthis figure, the reader is referred to the web version of this ar-ticle). Indeed if y> 1, the content of Gd would increase the ra-tio RA/RB over 1.37 and thus a Hydrogen InducedAmorphization (HIA) would occur according to Aoki andMasumoto [24], which could avoid a correct cyclability. Onthe contrary, with an extra amount of light element such asAl or Mg ( y< 1 and/or x> 0) it would result in a materiallighter than the parent AB2 compound and with an expectedlower hydrogen equilibrium pressure.
in 4a and 4c sites (exchange rate t¼ 50%) and (d) evolution of the intensities
of the first 2 peaks with t increasing from 0 to 50%.
3.2. Structural anomaly: disorder andexchange phenomenon
The ball-milling method is known to increase the latticestrains and decrease the crystallite sizes. For high ball-millingdurations, the product obtained is mainly amorphous and soa heat treatment is necessary to get a crystallised compound.Ball-milling can trigger a more important modification in theC15 crystal structure which is an exchange phenomenon be-tween Gd and Mg in both 4a and 4c sites. An exchange ratet is defined as
t¼ Occ:ðMg in 4aÞOcc:ðMg in 4aÞ þOcc:ðGd in 4aÞ�
Occ:ðMg in 4aÞ ¼ Occ:ðGd in 4cÞOcc:ðMg in 4aÞ þOcc:ðMg in 4cÞ ¼ 1
where Occ. stands for relative occupation, which allows us tocharacterise the amounts of Gd and Mg in 4a and 4c sites.Fig. 5 shows the theoretical XRD patterns as a function ofthe exchange rate.
With a 0.5/0.5 rate, the symmetry is improved and the com-pound adopts the MgCu2 C15 structure, triggering the extinc-tion of the (h00) diffraction peaks if h¼ 2n (as seen in Fig. 5dwith the (200) peak). We assumed that intermediate exchangerates between these two limits are possible. The last step ofour Rietveld refinements concerns the DebyeeWaller factor(Biso). If Gd and Mg occupy the 4a and 4c sites, respectively,a negative value of Biso for the 4a sites (meaning that the elec-tronic densities are too high) and a high Biso value for the 4csites (meaning that the electronic densities are too weak) areobtained. In other words the DebyeeWaller factor is consis-tent only if a balanced electronic distribution is possible be-tween the sites. Such an observation proves that Gd and Mgoccupy both the 4a and the 4c sites. For GdMgNi4 ball-milledand annealed, the exchange rate is about 14% (Fig. 6a).
Many parameters can influence this process like the heattreatment conditions, ball-milling duration, composition, etc.For example, in the case of GdMgNi4�xAlx, the increase ofAl content improves the steric hindrance effects which
Fig. 6. Rietveld refinement of GdMgNi4�xAlx obtained by Mechanical Alloying and subsequent heat treatment for (a) x¼ 0 (t z 14%) and (b) x¼ 0.5
(t z 7%).
183J.-G. Roquefere et al. / Intermetallics 16 (2008) 179e187
decrease the mobility of atoms within the lattice cell andthus it is obvious to obtain a lower t rate. In the case ofGdMgNi3.5Al0.5, the calculated exchange rate is t z 7% asillustrated in Fig. 6b.
It is worth pointing out that in the XRD pattern, a verysharp peak at 2q¼ 38� is the signature of a trick of experimentwhile the broad one at 31� is due to a GdNi5 impurity. As pre-viously mentioned, the synthesis conditions are also of primeimportance in this disorder phenomenon. Ball-milling en-hances GdeMg exchange but the resulting material is oftenmetastable and a long time heat treatment permits to decrease
the disorder rate. In the same approach, melted and annealedcompounds are more stable and this results in lower exchangerates as shown in Fig. 7.
Regarding the uncertainty in the exchange rate determina-tion by the mean of Rietveld refinements, the plot can be as-sumed to be linear. For both methods, t z 0% when the cellparameter reaches about 7.16 A (the highest cell parameterpossible in GdMgNi2.8Al1.2). The ideal C15b structure onlyexists when there is no REeMg exchange. Some other synthe-sis methods are currently investigated to have a null disorderrate for any composition.
6,90 6,95 7,00 7,05 7,10 7,15 7,200
5
10
15
20
diso
rd
er rate (%
)
a (Å)
milled samplesmelted samples
Fig. 7. Disorder rate as a function of the cell parameter in GdMgNi4�xAlx for
both milled and melted samples.
184 J.-G. Roquefere et al. / Intermetallics 16 (2008) 179e187
3.3. Magnetic properties: influence of the crystallinity
In RENi2 compounds (RE¼ Rare Earth element), a ferro-magnetic transition occurs at low temperature resulting to anindirect coupling between the RE atoms through the conduct-ing electrons. Such a coupling is known as RKKY interaction(RuddermanneKitteleKasuyaeYosida) [25]. Because of theweak radial density of the 4f electrons in rare earth elements,they cannot interact directly. The electrons belonging to theouter shells of the RE transfer to the nickel, occupying theNi 3d band. As a result the transition metal becomes non-magnetic and only the RE element is magnetic. GdMgNi4 isexpected to have the same behaviour since Mg is diamagnetic.In order to compare the magnetic behaviour of differentGdMgNi4 samples, the magnetizations were measured asa function of temperature (Fig. 8).
Each sample presents an increase of the magnetization atlow temperature which is due to the coupling of parallel elec-tronic spins in ferromagnetic compounds. The sample meltedand subsequently annealed (white squares) exhibiting a Curietemperature Tc¼ 77.6 K which is very close to the one mea-sured for GdNi2 (75.0 K; white circles) and to others availablein literature [26]. The saturating magnetization for that sampleis twice lower than the one measured for GdNi2 (i.e. 1.8 and3.6 mB/mol). Indeed the composition GdMgNi4 can be writtenas Gd0.5Mg0.5Ni2 and the magnetic moment of Gd can be con-sidered as diluted by a factor 2. This is again a proof for theexchange between 4a and 4c crystallographic sites (i.e. be-tween Gd and Mg). As no new crystallographic order exists,the magnetic behaviour is strictly comparable with the mothercompound GdNi2. If no exchange rate would exist, the mag-netic behaviour would not follow the dilution rule and evenwould not follow the same magnetic behaviour. However,for the sample ball-milled and annealed, the magnetization in-creases rather monotonously from 70 K. For the ball-milledsample, the magnetization is very low until 20 K. Anywaya slight transition is noticed at Tc¼ 73(1) K (inset in Fig. 8
corresponding to the Field Cooled and Zero Field Cooled).Such results are in rather good agreement with the previouscrystallographic observations. The ball-milled sample is al-most amorphous and so, no long range order exists. The de-crease (and the disappearance) of the ordering temperatureas a function of the decrease of the crystallinity has alreadybeen reported for several compounds. As an example, forGdNi2, the ball-milling process induces a decrease of theordering temperature from 77 to 73 K [26] as the compoundprepared by melt spinning exhibits a Curie temperature ofonly 38 K [27]. The hysteresis between the ZFC and FCcurves can be attributed to the loss of crystallinity inducedby the energetic ball-milling (magnetocrystalline anisotropy).Finally, the drastic increase of the magnetization below 15 Kis still unexplained.
In the RKKY model, the magnetic moment of nickel is nottaken into account. However, recent studies have shown thatNi is magnetic in such alloys ðmNi
eff ¼ 0:24 mBÞ and that itcan couple ferrimagnetically with Gd [28e30]. Such a conclu-sion is being currently investigated in GdMgNi4 within theMolecular Field Theory [31].
The susceptibility measurement under a field of 3 T of crys-talline GdMgNi4 prepared by ball-milling and subsequent heattreatment is presented in Fig. 9.
The curve follows CurieeWeiss law in such a field range,with a positive paramagnetic Curie temperature (qp¼25.5 K). The calculated effective magnetic moment is meff ¼7:72 mB which is very close to the theoretical value for Gd3þ
ði:e: mtheoreff ¼ 7:94 mBÞ. The linearity is no longer noticed below
150 K because of the combined effects of (i) the short rangeorder which increases as the temperature becomes closer tothe ordering temperature, (ii) the magnetic moment of Ni inthe compound sub-lattices [29,31] or (iii) the presence ofsome Ni clusters at the surface of the sample as frequentlyseen in such alloys.
3.4. Hydrogen sorption properties
Cubic AB2 Laves phases can host H atoms in some tetrahe-drons [A2B2]. These sites only indeed satisfy the criteria ofWestlake and Switendick. The compounds REMgNi4 havevery different behaviours towards hydrogen depending onthe rare earth element [12]. For a correct cyclability, Gd seemsto be the most appropriate (i.e. no orthorhombic distortion[14], no HIA [32], no decomposition [33]). Isotherm PCicurves are presented in Fig. 10.
The two samples milled and annealed for 3 days exhibita wide equilibrium pressure plateau which is the signatureof a good cyclability and a large reversibility range. In thecase of Gd0.8Mg1.2Ni4, a maximum uptake of 0.55 H/M ismeasured (i.e. Gd0.8Mg1.2Ni4H3.3 meaning a reversible capac-ity of about 1.0 wt.%). The estimated hydriding enthalpy (e.g.DHhydr. z�30 kJ/mol H2 is close to the one reported forequivalent compounds (i.e. �31.4 kJ/mol H2 reported by Ter-ash*ta and Akiba for GdMgNi4 [17]). The annealed compoundGd1.4Mg0.6Ni4 almost exhibits the same properties but it’sworth pointing out that in spite of a ratio RA/RB higher than
0 10 20 30 40 50 60 70 80 90 100 110 120 1300
1
2
3
10.0 K
0 20 40 60 80 100 1200,0
0,1
0,2
0,3
0,4
M (μ
B/m
ol)
Temperature (K)
77.6(1) K
GdNi2
GdNi4Mg melted + annealed
GdNi4Mg ball-milled + annealed
GdNi4Mg ball-milled
Mag
netizatio
n (μ
B/m
ol)
Temperature (K)
75.0(1) K
73(1) K
Fig. 8. Magnetization as a function of the temperature of GdMgNi4 for different synthesis methods (a zoom of the curve related to the milled sample is presented in
the inset).
185J.-G. Roquefere et al. / Intermetallics 16 (2008) 179e187
1.37, hydrogen induced amorphization is not observed, proba-bly because of the exchange of Gd and Mg on the 4a and 4ccrystallographic sites.
The two samples ball-milled have a different behaviour.Since they are metastable, they absorb hydrogen out of equi-librium and thus no plateau is noticed. The desorption is neverfully achieved whatever be the temperature and the sample. Ithas been shown previously that amorphous Mg-based alloyshad better kinetics and absorption capacities than crystallisedones; however, the sorption reversibility has never been no-ticed in amorphous samples. Finally, it is worth pointing out
0 50 100 150 200 250 3000
5
10
15
20
25
30
35
40
Cm = 7.45µeff = 7.72 µB/Gdθp = 25.5 K
GdNi4Mg µ0H = 3 T
m
-1 (m
ol/em
u)
Temperature (K)
Fig. 9. Reverse magnetic susceptibility in GdMgNi4 (dot line represents a lin-
ear fit).
that the full desorption was never reached at the end of eachPCi experiment. Then, after each PCi experiment, a manualdesorption under dynamic vacuum was done so that the nextcurve is probably shifted (i.e. the initial value is differentfrom 0).
3.5. Electrochemical performances
Some NieMH batteries were charged and discharged forseveral cycles and the capacities were measured during thedischarge. Fig. 11 presents the electrochemical capacity dur-ing cycling as well as the number of exchanged hydrogenatoms.
The crystallised compound provides a capacity Q¼ 70mA h/g after 10 cycles (about one hydrogen atom exchanged).Then the capacity decreases almost linearly. The amorphouscompound has a very weak capacity in comparison with thefirst sample. This result is quite different from those ofWang et al. [15,34] who reported the electrochemical effi-ciency of amorphous alloys. Actually, they studied the capac-ities of some Mg-based alloys as a function of the particlesizes and sintering temperatures rather than the crystallinityand so, it is obvious that small particles have a larger reactivesurface. As stated in previous paragraph, a high crystallinity isnecessary to have a complete reversibility towards hydrogen,which is the most important parameter for a good cyclability.These results are in good agreement with the previous resultsobtained on solidegas reaction, however, the difference in themaximum hydrogen uptake is still unexplained.
0,0 0,1 0,2 0,3 0,4 0,5 0,6
0,01
0,1
1
PH
2 (M
Pa)
0,01
0,1
1
PH
2 (M
Pa)
0,01
0,1
1
PH
2 (M
Pa)
0,01
0,1
1P
H2 (M
Pa)
H/M
Absorption 243KDesorption 273KDesorption 243K
Gd0,8
Mg1,2
Ni4 ball milled and annealed at 680°C for 3 days
0,0 0,1 0,2 0,3 0,4 0,5
Gd1,4
Mg0,6
Ni4
ball milled and annealed at 680°C for 3 days
H/M
Desorption 273KAbsorption 243KDesorption 243K
0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1,0
Gd0,8
Mg1,2
Ni4 ball milled
H/M
Absorption 273KDesorption 273KAbsorption 243KDesorption 243K
0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9
Gd1,4
Mg0,6
Ni4
ball milled
H/M
Absorption 303KDesorption 303KAbsorption 239KDesorption 239K
Fig. 10. PCi curves for Gd0.8Mg1.2Ni4 (left part) and Gd1.4Mg0.6Ni4 (right part).
186 J.-G. Roquefere et al. / Intermetallics 16 (2008) 179e187
4. Conclusions
The compounds GdyMg2�yNi4�xAlx with 0.4< y< 2 and0< x< 1.2 have been successfully elaborated by both milling
0 10 20 30 40 50 60 7010
20
30
40
50
60
70
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1,0
1,1
GdNi4Mg crystallisedGdNi4Mg amorphous
Q (m
Ah
/g
)
Numbers of cycles
H / f.u
.
Fig. 11. Electrochemical capacity and number of exchanged hydrogen during
cycling.
and melting methods. The structural properties of thesecompounds are strongly influenced by the synthesis routeand a local disorder as well as an exchange phenomenon iscreated and enhanced by energetic ball-milling. Magneticmeasurements confirmed such a disorder in comparison withmelted samples. A long heat treatment is therefore necessaryto obtain a stable and crystallised material with the idealC15b structure (MgCu4Sn-type).
Hydrogen sorption properties highly depend on these struc-tural properties. To provide a good cyclability, high hydrogenuptakes/electrochemical capacities, and ambient usage condi-tions, the material must be stable and well crystallised. Fur-thermore, the addition of Al instead of Ni and/or Mg insteadof Gd permits to prevent an eventual HIA and increases theweight capacity.
Finally, materials such as Gd0.6Mg1.4Ni3Al, for example,are some serious candidates for hydrogen storage.
Acknowledgements
The support from New Energy and Industrial TechnologyDevelopment Organization (NEDO), Japan via a subcontractbetween AIST and ICMCB is largely appreciated.
187J.-G. Roquefere et al. / Intermetallics 16 (2008) 179e187
References
[1] Li H, Eddouadi M, O’Keeffe M, Yaghi OM. Nature 1999;402:276e9.
[2] Bouaricha S, Dodelet JP, Guay D, Huot J, Schulz R. J Alloys Compd
2001;325:245e51.
[3] Chen P, Xiong Z, Luo J, Lin J. Nature 2002;420:302e5.
[4] Bobet J-L, Castro FJ, Chevalier B. J Alloys Compd 2005;376(1e2):
245e8.
[5] Guo J, Wei W, Ma S, Gao Y. Mater Sci Eng B 2003;98:21e4.
[6] Bobet J-L, Desmoulins-Krawiec S, Grigorova E, Cansell F, Chevalier B.
J Alloys Compd 2003;351:217e21.
[7] Barkhordarian G, Klassen T, Bormann R. J Alloys Compd 2004;364:
242e6.
[8] Tanaka S, Clewley JD, Flanagan TB. J Phys Chem 1977;81(17):1684e8.
[9] Sai Raman S, Srivastava ON. J Alloys Compd 1996;241:167e74.
[10] Jacob I, Shaltiel D. J Less-Common Met 1979;65:117e28.
[11] Reilly JJ, Wiswall Jr RH. Inorg Chem 1968;7:2254e6.
[12] Bobet J-L, Lesportes P, Roquefere J-G, Chevalier B, Asano K, Sakaki K,
et al. Int J Hydrogen Energy 2007;32(13):2422e8.
[13] Prigent J, Gupta M. J Alloys Compd 2007;446e447:90e5.
[14] Guenee L, Favre-Nicolin V, Yvon K. J Alloys Compd 2003;348:129e37.
[15] Wang ZM, Zhou H, Zou RP, Yao Q. J Alloys Compd 2007;429:260e3.
[16] Fang S, Zhou Z, Zhang J, Yao M, Feng F, Northwood DO. J Alloys
Compd 1999;293e295:10e3.
[17] Terash*ta N, Akiba E. Mater Trans 2006;47(8):1890e3.
[18] Rodriguez-Carvajal J. FULLPROF: a program for Rietveld refinement
and pattern matching analysis. Abstracts of the satellite meeting on
powder diffraction of the XV congress of the IUCr, Toulouse, France,
1990. p. 127.
[19] Schulz R, Boily S, Huot J. Canadian Patent, Ser.-Nr. 2207149, 1999.
[20] Bobet J-L, Pechev S, Chevalier B, Darriet B. J Alloys Compd
1998;267:136e41.
[21] Manfrinetti P, Gschneidner Jr KA. J Less-Common Met 1986;123:
267e75.
[22] Oesterreicher H, Bittner H. J Less-Common Met 1980;73:339e44.
[23] De Negri S, Giovannini M, Saccone A. J Alloys Compd 2005;
397(1ndash2):126e34.
[24] Aoki K, Masumoto T. J Alloys Compd 1993;194:251e61.
[25] Ruderman MA, Kittel C. Phys Rev 1954;96:99e102.
[26] Chevalier B, Bobet J-L, Nakhl M, Etourneau J. J Alloys Compd 2001;
320:33e9.
[27] Lee K, Heiman N. AIP Conf Proc 1974;24:108.
[28] Yano K, Umehara I, Miyazawa T, Adachi Y, Sato K. Physica B 2005;
367:81e5.
[29] Yano K, Tanaka Y, Matsumoto I, Umehara I, Sato K, Adachi H, et al. J
Phys Condens Matter 2006;18:6891e5.
[30] Paulose PL, Patil S, Mallik R, Sampathkumaran EV, Nagarajan V. Phys-
ica B 1996;223&224:382e4.
[31] Yano K. J Magn Magn Mater 2000;208:207e16.
[32] Aono K, Orimo S, Fujii H. J Alloys Compd 2000;309:L1e4.
[33] Bobet J-L, Grigorova E, Chevalier B, Khrussanova M, Peshev P. Interme-
tallics 2006;14:208e12.
[34] Xu X, Zhou HY, Zou RP, Zhang SL, Wang ZM. J Alloys Compd
2005;396:247e50.
