Available online at www.sciencedirect.com

www.elsevier.com/locate/actamat

Acta Materialia 59 (2011) 5821–5831

Feasibility and performance of the mixture of MgH2 and LiNH2 (1:1)as a hydrogen-storage material

J.J. Hu ⇑, E. Rohm, M. Fichtner

Institute of Nanotechnology, Karlsruhe Institute of Technology (KIT), P.O. Box 3640, D-76021 Karlsruhe, Germany

Received 23 March 2011; received in revised form 17 May 2011; accepted 29 May 2011Available online 24 June 2011

Abstract

A 1:1 molar ratio mixture of MgH2 and LiNH2 was predicted to release 8.1 wt.% H under moderate conditions. This binary mixtureof MgH2–LiNH2 was found to be a multinary complex system induced by mechanical ball milling, due to the metathesis reaction betweenthe initial components. It was found that dehydrogenation from this system was initialized by the formation of LiH and Mg(NH2)2 viasuch metathesis. The hydrogen sorption performance studied in this work shows a strong influence of the ball-milling parameters whichdetermine the subsequent dehydrogenation pathways. An adequate ball milling facilitates hydrogen release, whereas insufficient millingresults in a sluggish hydrogen desorption and severe NH3 emission. A maximum amount of 7.3% was obtained with formation of theternary nitride LiMgN; however, desorption temperatures of up to 600 �C had to be applied.� 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Hydrogen storage; Thermodynamics; Metal amides; Hydrogen desorption

1. Introduction

Reversible hydrogen storage systems with high H-capac-ity are crucial to onboard vehicle applications. Althoughthe US Deparment of Energy’s 2015 performance targetwas revised down to a system gravimetric density of5.5 wt.% instead of 9.0 wt.%, this still poses a greatchallenge for materials investigators: a chemical hydrogenstorage material has to contain more than 7 mass% hydro-gen which can be endothermically released at temperatureslower than 200 �C [1].

A Li–Mg–N–H system comprising a 1:1 molar ratio ofLiNH2–MgH2 has been selected for further investigation[2]. Using first-principles calculations, Alapati et al. [3] pre-dicted that the amide–hydride combination of LiNH2 andMgH2 at a 1:1 molar ratio would desorb 8.1 wt.% H viathe following reaction:

LiNH2 þMgH2 ! LiMgNþ 2H2 ð1Þ

1359-6454/$36.00 � 2011 Acta Materialia Inc. Published by Elsevier Ltd. All

doi:10.1016/j.actamat.2011.05.059

⇑ Corresponding author. Tel.: +49 72160828915; fax: +49 72160826368.E-mail address: jianjiang.hu@kit.edu (J.J. Hu).

An enthalpy change of 29 kJ/mol H2 was also calculatedby means of density functional theory (DFT); this changefalls in the range of ideal thermodynamics for a hydrogen-storage material. Following this prediction, Lu et al. [4]demonstrated experimentally an excellent agreement withthe above prediction. A capacity of 8.1 wt.% was measuredin the temperature range of 160–220 �C with LiMgN asdehydrogenation product. The desorbed product LiMgNdoped with TiCl3 could fully be rehydrogenated. In a reportby Osborn et al. [5], however, only 3.4 mass% hydrogenrelease was measured at 210 �C. Moreover, a large amountof NH3 emission was detected by thermogravimetry and noor little LiMgN was formed. In another attempt by Liuet al. [6], a total of 6.1 wt.% H2 was obtained from theLiNH2–MgH2 system in two steps at 222 and 390 �C,respectively. The enthalpy change of dehydrogenation wasfound to be 45.9 kJ/mol H2, which is much higher thanthe predicted value. Moreover, Mg3N2 was identified asthe desorbed product instead of LiMgN. Different path-ways and reaction products resulted due to various samplepreparation conditions [7], reported by Liang et al. in asupplementary work to Ref. [6]. In a recent follow-up

rights reserved.

5822 J.J. Hu et al. / Acta Materialia 59 (2011) 5821–5831

communication by Lu et al. [8], the authors also found thatthe reaction pathway of the LiNH2–MgH2 mixture wasdependent on the mechanochemical processing conditions.Mild ball-milling conditions would favor the formation ofLiMgN, i.e. the predicted dehydrogenation product.

The diverse performances of LiNH2–MgH2 found bydifferent investigators indicate the complexity of thishydrogen-storage system. In order to investigate the feasi-bility and performance in relation to the sample processing,different ball-milling parameters, such as revolutions perminute (rpm) and duration, were tested with or withoutTiCl3 as additive. Sampling at different desorption stageswas performed to trace the phase and structure develop-ment during dehydrogenation.

2. Experimental

2.1. Sample preparation

LiNH2 (Sigma Aldrich, >95%), MgH2 (Alfa Aesar,98%) and TiCl3 (Acros Organics, 99%) were stored in aglovebox under Ar as a protecting atmosphere and usedas received. For the ball-milling process, the starting chem-icals were loaded into a milling vessel inside the glovebox,at a 1:1 molar ratio of LiNH2:MgH2, with varying amountsof TiCl3. Ball milling was performed using a Retsch PM400E. While keeping the powder to ball weight ratio con-stant at 1:90, revolution rates of 100, 150, 200 and400 rpm were employed. Milling periods were varied from2.5 to 96 h. TiCl3 was used to investigate its effect onhydrogen sorption. Detailed ball-milling conditions arelisted in Table 1, which covers various combinations ofmilling energy and duration.

Fourier transform infrared spectroscopy (FTIR) mea-surements were conducted using a Perkin Elmer SpectrumGX FTIR system. The powdery samples were mixed withKBr and pressed into pellets and measured at a resolutionof 4 cm�1.

Table 1Sample preparation conditions.

Sample name Milling conditions TiCl3 addition

S100 rpm-2.5 h 100 rpm, 2.5 h –S100 rpm-2.5 h-Ti 100 rpm, 2.5 h 5 wt.% TiCl3S100 rpm-48 h 100 rpm, 48 h –S100 rpm-48 h-Ti 100 rpm, 48 h 10 wt.% TiCl3S100 rpm-96 h 100 rpm, 96 h –S100 rpm-96 h-Ti 100 rpm, 96 h 10 wt.% TiCl3S150 rpm-10 h 150 rpm, 10 h –S150 rpm-15 h 150 rpm, 15 h –S150 rpm-20 h 150 rpm, 20 h –S150 rpm-25 h 150 rpm, 25 h –S150 rpm-30 h 150 rpm, 30 h –S200 rpm-5 h 200 rpm, 5 h –S200 rpm-10 h 200 rpm, 10 h –S200 rpm-10 h-Ti 200 rpm, 10 h 5 wt.% TiCl3S200 rpm-20 h 200 rpm, 20 h –S400 rpm-20 h-Ti 400 rpm, 20 h 16 wt.% TiCl3

Phase identification was performed by X-ray diffractom-etry (XRD) on a Philips X’PERT diffractometer (Cu Karadiation). In order to protect samples against air, a sam-ple-holder consisting of a Kapton foil hood and a siliconsingle-crystal base was used. The powdery samples werespread evenly onto the silicon crystal and then sealed withthe Kapton foil hood inside the glovebox. The measure-ment was conducted in the 2h range of 10–80� at a steplength of 0.02�.

Differential scanning calorimetry (DSC) measurementswere performed on a Netzsch DSC 204 HP housed insidethe glovebox to detect heat effects accompanying hydrogendesorption. Samples were heated at 5 �C min�1 under 3 barHe.

For the quantification of H desorption, a carefully cali-brated homemade Sieverts system was used to measure theH2 volume evolved in a temperature programmed desorp-tion (TPD) mode or isothermal desorption mode.The TPD mode was run at a temperature ramping of5 �C min�1 from 30 to 600 �C and held at 600 �C for 1 h,whereas for the desorption isotherms the temperature wasfirst raised to 225 �C at 5 K min�1 and held at this temper-ature for at least 20 h.

The N and H contents were determined by elementalanalysis using an instrument from Elementar Analysensys-teme GmbH.

3. Results and discussion

3.1. Metathesis reaction through ball milling

The amide–hydride combination of LiNH2–MgH2 at2:1 molar ratio was first investigated by Luo [9] as a hydro-gen-storage system. A considerable improvement in hydro-gen-sorption properties was achieved compared to itspredecessor LiNH2–LiH system [10]. Interestingly, aftercycling of the system, Mg(NH2)2 and LiH were formed,instead of LiNH2 and MgH2. It was later recognized thatthere exists a metathesis conversion from LiNH2 andMgH2 to Mg(NH2)2 and LiH that could be achieved byheating LiNH2 and MgH2 at 200 �C under H2 pressure[11].

2LiNH2 þMgH2 !MgðNH2Þ2 þ 2LiH ð2ÞIn the presence of LiBH4, this conversion takes place at

lower temperature [12].FTIR spectra are shown in Fig. 1 after milling at

100 rpm for various durations. For the shortest millingtime of 2.5 h, the characteristic N–H vibrations of the pri-mary component LiNH2 at 3312 and 3258 cm�1 wereclearly detected for the samples with and without TiCl3addition. Differences appear as milling was extended to48 h. While the doublet of 3312 and 3258 cm�1 persists inthe pure LiNH2–MgH2 sample (S100 rpm-48 h), broaden-ing and shifting of the absorbance become obvious withthe TiCl3-doped sample (S100 rpm-48 h-Ti), indicatingthe formation of Mg(NH2)2 at 3272 and 3326 cm�1. With

Fig. 1. FTIR spectra of samples ball milled at 100 rpm.

J.J. Hu et al. / Acta Materialia 59 (2011) 5821–5831 5823

further milling to 96 h, the N–H stretching from LiNH2

was further weakened. In S100 rpm-96 h-Ti, the formationof Mg(NH2)2 was also detected, as in S100 rpm-48 h-Ti.This suggests that the conversion reaction (2) was takingplace during ball milling. Compared with FTIR spectraof the samples prepared with the same milling durations,it seems that the addition of TiCl3 could promote themetathesis conversion to Mg(NH2)2–LiH. The gradualconversion was verified clearly at 150 rpm, i.e. with highermilling energy (Fig. 2). The shape of the N–H absorbancein LiNH2 started to change after milling for 20 h. Mixedfeatures of the N–H vibrations originated from LiNH2

and Mg(NH2)2, respectively, are discernible for the samplemilled for 30 h. Liu et al. [6] found that Mg(NH2)2 was firstformed at a ball milling time of 12 h and consumed after36 h, whereas Mg(NH2)2 could already be detected after1 h milling by Lu [8] using a vibrational ball mill. The for-

Fig. 2. FTIR spectra of samples ball milled at 150 rpm.

mation of Mg(NH2)2, i.e. the metathesis reaction, seemsdependent on milling energies. Generally, high millingenergy would shorten the metathesis conversion process.However, as soon as Mg(NH2)2 and LiH became availablein the mixture, further reactions can be initialized, whichwill be described in the following sections.

3.2. Thermal behavior in dependence of ball-milling

conditions

Endothermic dehydrogenation is a prerequisite for areversible hydrogen-storage system. Hence, the thermalbehavior is of critical significance for hydrogen storage.The DSC curves in the temperature range between 100and 500 �C reveal very complex thermal behaviorsthat are strongly related to the milling parameters (Figs.3 and 4). For the sample series at 150 rpm, an exothermicpeak at around 200 �C was detected for the samplesS150 rpm-5 h and S150 rpm-10 h. With increased ball-mill-ing durations, the exothermic heat decreased gradually andan endothermic peak appeared at about 190 �C from 15 honwards. For milling times longer than 15 h, several endo-thermic peaks appear in the temperature range investi-gated. In a similar way, the sample series at 100 rpmcomprised of multiple thermal processes, with an exother-mic peak at about 200 �C for the sample S100 rpm-48 h,which was replaced by an endothermic peak at around190 �C with TiCl3 addition or extended ball milling of 96 h.

In order to find the origin of the exothermic heat effect,we carried out a heat treatment of the sample S200 rpm-5 h, in which insignificant metathesis was caused by mill-ing. The heat treatment was conducted at 5 �C min�1 andinterrupted at 236 �C, where the exothermic process ended(Fig. 5). In the FTIR spectra, absorbance peaks at 3165,3194 and 3272 cm�1 appeared in addition to the absor-bance at 3258 and 3312 cm�1 that originate from LiNH2

(Fig. 6). The absorbance at 3272 cm�1, albeit low in inten-

Fig. 3. DSC curves from the as-prepared samples ball milled at 150 rpmfor 10, 15, 20, 25 and 30 h.

Fig. 4. DSC curves from the as-prepared samples milled at 100 rpm for 48and 96 h with and without TiCl3.

Fig. 5. DSC curve of the as-prepared S200 rpm-5 h.

Fig. 6. FTIR spectra of S200 rpm-5 h: as-milled and heated to 236 �C.

Fig. 7. XRD profiles of S200 rpm-5 h: as-milled and heated to 236 �C.

5824 J.J. Hu et al. / Acta Materialia 59 (2011) 5821–5831

sity, is an indication of the formation of Mg(NH2)2, whichis supported by the XRD profile showing diffractions at the2h values of 14.9, 19.6, 23.3 characteristic of Mg(NH2)2

(01-072-0786, I41/acd) (Fig. 7). It is obvious that themetathesis took place readily upon heating the primarycomponents LiNH2 and MgH2.

Using thermodynamic data found in the literature [13],the reaction enthalpy of the metathesis (2) leading to theformation of Mg(NH2)2 can be obtained as follows:

DH h ¼ Df H hMgðNH2Þ2 þ 2Df H h

LiH � Df H hMgH2

� 2Df H hLiNH2

¼ ð�353Þ þ 2 � ð�91Þ � ð�75Þ � 2 � ð�182Þ¼ �278 kJ mol�1

The calculation result demonstrates that the metathesisreaction is an exothermic process. Thus, the exothermicevent observed by DSC was caused by the metathesis

reaction. Along with the FTIR results, we demonstratedthat the metathesis conversion can be realized either bymechanical milling or by heating. Thus, the various ther-mal effects ranging from exothermic to endothermic inFigs. 3 and 4 were related to the different extents of themetathesis attained in the ball-milling processing.

The additional vibrations at 3165 and 3194 cm�1

observed from S200 rpm-5 h heated to 236 �C (Fig. 6) aresigns of the N–H absorbance of imide groups of Li2NHand MgNH, respectively, which were recognizable fromthe XRD diffractions of Li2NH and cubic modification ofLi2Mg(NH)2 [14]. This means that during the exothermicevent not only did the metathesis reaction occur, but dehy-drogenation from the in situ formed Mg(NH2)2 and LiHalso took place at temperatures above 200 �C. On the otherhand, the consumption of Mg(NH2)2 formed from themetathesis may explain the low intensity at 3274 cm�1 of

Table 2Crystallite size of MgH2 estimated by the Scherrer method.

Sample Crystallite size (nm)

S100 rpm-48 h 10.5S100 rpm-48 h-TiCl3 5.4S100 rpm-96 h 3.9S100 rpm-96 h-TiCl3 3.0

J.J. Hu et al. / Acta Materialia 59 (2011) 5821–5831 5825

Mg(NH2)2 coupled with the appearance of vibrations fromthe imide groups.

3.3. Influence of ball-milling conditions on composition and

phase development

The XRD profiles of the samples milled at 100 rpm for2.5 h exhibit strong and narrow diffraction peaks of bothstarting chemicals (Fig. 8), suggesting that the crystalliteswere not greatly reduced under the ball-milling conditions.The presence of TiCl3 was nevertheless not detected, pre-sumably due to its low content. The metal magnesiumwas an impurity which originated from MgH2. Broadeningof the diffraction peaks was observed in samples ball milledwith higher energies and for longer times (Fig. 9). Crystal-lites were clearly crushed by ball milling for 48 and 96 h.

Fig. 8. XRD profiles of briefly ball milled (100 rpm, 2.5 h) samples withand without TiCl3.

Fig. 9. XRD profiles of the as-prepared samples milled at 100 rpm for 48and 96 h with and without TiCl3.

The diffraction intensities were reduced significantly. Aftermilling for 96 h, diffraction peaks from LiNH2 disappearcompletely. Using the Scherrer calculator integrated inthe X’Pert HighScore software with commercial MgH2

(purity 98%) as standard provided an estimate of the crys-tallite size of MgH2 (Table 2). Although the crystallite sizeof MgH2 can be reduced down to a few nanometers, specialmeans have to be employed, such as ultra-high-energy,high-pressure mechanical milling [15] or using reaction buf-fer to separate MgH2 particles [16]. It is notable that thecrystallite size of MgH2 in this study was reduced to thenanometer scale with low-energy milling in the presenceof a “soft” material, i.e. LiNH2. We believe that themetathesis conversion or consumption of MgH2 duringmilling played a more important role. Due to the highdegrees of amorphicity, deviation in the estimation can cer-tainly arise. However, from the extremely low intensities ofthe diffraction, it is quite possible that the crystallite sizeswere in the lower nanometer range. The addition of TiCl3facilitated a reduction in crystallite size under the samemilling conditions.

The sample series at 150 rpm shows a rougher back-ground compared to the series at 100 rpm in the progressof crystal phases over 5–30 h of milling (Fig. 10). The dif-fraction intensities of LiNH2 at 2h = 30.5� and 50.7�decreased continuously, indicating an increase in amorph-icity. In contrast to the FTIR observation, no diffractionfrom Mg(NH2)2 was detected, which was probably present

Fig. 10. XRD profiles of the as-prepared samples without TiCl3 additionmilled at 150 rpm for 10, 15, 20, 25 and 30 h, respectively.

Fig. 11. Dehydrogenation isotherms at 225 �C for samples prepared underdifferent conditions.

Table 4Possible reactions in the 1:1 mixture of LiNH2 and MgH2.

Reactions DH, kJ/mol H2

LiH + LiNH2! Li2NH + H2 (3) 45 [10]2LiH + Mg(NH2)2! Li2Mg(NH)2 + 2H2 (4) 40 [18]2Mg(NH2)2 + 2LiH! Li2Mg2(NH)3 + NH3 + 2H2 (5) NAMgH2 + Mg(NH2)2! 2MgNH + 2H2 (6) NA [19]2MgH2 + Mg(NH2)2!Mg3N2 + 4H2 (7) 3.5 [20]

5826 J.J. Hu et al. / Acta Materialia 59 (2011) 5821–5831

as an amorphous phase, since Mg(NH2)2 becomes amor-phous readily upon milling. Nonetheless, two weak peaksat 38� and 44.2�, which are assignable to LiH, became pro-gressively stronger with increased milling time, an indica-tion of the metathesis reaction detected by FTIR.

3.4. Influence of ball-milling conditions on dehydrogenation

performance

In order to determine the dehydrogenation quantita-tively, we chose a volumetric method instead of thermo-gravimetric means. For nitrogen-containing hydrogen-storage systems, the thermogravimetric method wouldincrease the error in the measurement error of hydrogenamount in the event of NH3 generation. H amounts higherthan theoretical values were reported in the literature dueto NH3 emission [5,8]. Conversely, the contribution ofNH3 to the partial volume is only 1/8.5 of H2 at the sameweight. Therefore, the amount of H2 release measuredusing volumetric methods is much less affected by NH3

emission.Pure LiNH2 decomposes thermally to Li2NH and NH3,

which is strongly inhibited by the NH3 equilibrium pres-sure [17]. The onset decomposition temperature lies above300 �C. The other component MgH2 releases H2 at temper-atures higher than 300 �C. We first performed dehydroge-nation tests at a moderate temperature, 225 �C, at whichneither of the pure starting components is able to dehydro-genate. The hydrogen amounts released in isothermdesorption mode at 225 �C and TPD at 600 �C are listedin Table 3.

Surprisingly, 3.4% H2 was obtained at 225 �C from thebriefly ball-milled samples (S100 rpm-2.5 h), though witha very sluggish kinetics (Table 3 and Fig. 11). This dehy-drogenation performance is close to the result obtainedby Osborn et al. [5] in an isothermal volumetric measure-ment at 210 �C. As the applied temperature of 225 �C isconsiderably lower than the individual decomposition tem-peratures of LiNH2 and MgH2, we attribute the measuredH amount of 3.4% to the interaction between LiNH2 andMgH2. As found above from the FTIR spectra and XRDprofiles, the metathesis reaction between LiNH2 andMgH2 could be facilitated by milling or heating. With the

Table 3Dependence of hydrogen-release performance on ball-milling conditions.

Sample Isothermal H2

desorption at225 �C in thefirst 5 h (wt.%)

Isothermal H2

desorption at225 �C (wt.%)

Total H2

desorption,TPD 600 �C(wt.%)

S100 rpm-2.5 h 1.6 3.4 5.9S100 rpm-2.5 h-Ti 1.6 3.3 6.0S100 rpm-48 h 3.3 3.7 6.8S100 rpm-48 h-Ti 3.8 4.5 6.3S100 rpm-96 h 4.1 5.0 7.2S100 rpm-96 h-Ti 3.6 3.9 5.6S200 rpm-10 h 3.9 4.9 7.3

metathesis taking place, the originally binary system con-sisting of LiNH2 and MgH2 became a quaternary one com-prising of LiH, MgH2, LiNH2 and Mg(NH2)2. Amongthese chemical species, a variety of reactions leading toH2 release are known in the literature (Table 4) [10,18–20].

All the reactions possess moderate dehydrogenationthermodynamics. Therefore, the metathesis reaction mayfunction as an initializing step for the hydrogen releasefrom the mixture of LiNH2 and MgH2.

Since little or no metathesis conversion was induced byball milling in the briefly milled samples, MgH2 and LiNH2

had first to overcome an extra barrier of metathesis viaheating at 225 �C, which slows down the kinetics. Additionof TiCl3 improved the kinetics to some extent, albeit with-out increasing the desorbed amount. Only 3.4 and 3.3 wt.%H2 were released in a period of 30 h, respectively. In con-trast, a reasonably fast kinetics was observed at 225 �Cfor samples with extended ball milling and/or addition ofTiCl3, whereby the metathesis step was partly or almostcompleted during the milling process. The major part ofhydrogen release occurred within the first 2–3 h, in contrastto the briefly milled samples. The dehydrogenation experi-enced a huge decrease in the kinetics after fast dehydroge-nation at the beginning, indicating an obvious change inthe dehydrogenation reaction. The highest value was mea-sured for the sample S100 rpm-96 h with 5.0 wt.%, close tothat of S200 rpm-10 h (4.9 wt.%), which is, however, muchlower than the predicted 8.1%. It should be noted thatunder the conditions of 96 h milling plus TiCl3 the hydro-

Table 5Elemental analysis results.

Sample Treatment H desorbed(%)

H content(%)

N content(%)

J.J. Hu et al. / Acta Materialia 59 (2011) 5821–5831 5827

gen release occurred during milling treatment and conse-quently a lower H2 desorption amount was measured.

The obvious discrepancy between the 5 wt.% hydrogenactually released and the predicted 8.1 wt.% at moderatetemperatures motivated us to investigate the system athigher temperatures as well. Dehydrogenation was per-formed under TPD mode with a temperature ramp of5 K min�1 from room temperature to 600 �C followed byholding this temperature for 1 h (Fig. 12). The onset tem-peratures for dehydrogenation was shifted to lower temper-atures by about 50 �C, compared to the briefly ball-milledsamples (S100 rpm-2.5 h and S100 rpm-2.5 h-Ti). Sufficientmilling with or without addition of TiCl3 obviously favoreddehydrogenation, as in the case of desorption at 225 �C.The samples with TiCl3 addition released less H2 than theircounterparts without TiCl3, a trade-off of the kinetic gainseen also in the isothermal measurements. In addition,the complex heat effects revealed by DSC are well reflectedin the TPD curves. Except for the samples involving exo-thermic metathesis reaction (briefly milled samples), thereare four stages of hydrogen desorption, each correspondingto the endothermic peaks in DSC curves, inferring that theevents are associated with hydrogen release. Thus, thedehydrogenation from the 1:1 system was found to be amultistep process, contradicting the one-step dehydrogena-tion claimed by Lu et al. [4]. For the samples which deliv-ered the highest H amount (S100 rpm-96 h and S200 rpm-10 h), the first desorption stage ended at about 270 �C (firstinflexion), corresponding roughly to the first endothermicpeak in DSC. This confirms that the first stage of dehydro-genation has its origins in reaction (4), which gives a calcu-lated H amount of 4.1% for the 1:1 system. At theinflexion, 3.7% H was desorbed, coinciding with theamount for which the desorption kinetics abruptlydecreased in the isothermal measurement (Fig. 11). It isclear from the TPD results that a considerable proportion

Fig. 12. Temperature programmed dehydrogenation at 5 �C min�1 fromroom temperature to 600 �C for samples prepared under differentconditions.

of the hydrogen could only be released at highertemperatures.

The highest amount of 7.3% H was measured at the tem-perature of 600 �C for the samples S200 rpm-10 h andS100 rpm-96 h, which also delivered the highest H2

amounts in desorption isotherms at 225 �C. The lowestvalue was found for the sample with the shortest milling.However, the highest value is considerably lower than thetheoretical quantity of 8.1%.

The N and H contents were examined by elemental anal-ysis on the starting chemicals and the typical samples with-out dopant TiCl3 (Table 5). Based on the results of thisanalysis, the purities of LiNH2 and MgH2 in terms of Hcontent were calculated to be about 96% and 93%, respec-tively. This means the H amount available in the 1:1 mix-ture is about 7.79% due to the impurities. While the Hcontent found in the sample by simply mixing LiNH2

and MgH2 with a mortar and pestle was 7.66%, which isclose to 7.79%, the H content measured in the samples pre-pared by ball milling decreased with ball-milling intensity,signaling a possible H loss during ball milling.

For the briefly ball-milled sample, the H contentsremaining after desorption at 225 and 600 �C were 4.03%and 1.22%, respectively. In contrast, the remaining Hamounts in the samples S200 rpm-10 h and S100 rpm-96 h were much lower. There was �2.5% H remaining afterisothermal desorption at 225 �C, and only traces of H weredetermined for desorption at 600 �C.

The relative N content is expected to increase with Hevolution as dehydrogenation progresses. In case of con-comitant NH3 generation, the N content decreases. For

LiNH2 As-received 0 8.44 58.47MgH2 As-received 0 7.13 0LiNH2–

MgH2

Calculated 0 7.79 27.28

Simplemixture

Mortar andpestle

0 7.66 27.63

S100 rpm-2.5 h

As-milled 0 7.46 27.07

S200 rpm-10 h

As-milled 0 7.24 27.47

S100 rpm-96 h

As-milled 0 7.28 27.37

S100 rpm-2.5 h

Desorbed at225 �C

3.3 4.03 26.80

S200 rpm-10 h

Desorbed at225 �C

4.9 2.58 28.55

S100 rpm-96 h

Desorbed at225 �C

5.0 2.40 28.85

S100 rpm-2.5 h

Desorbed at600 �C

5.9 1.22 24.56

S200 rpm-10 h

Desorbed at600 �C

7.3 0.41 30.00

S100 rpm-96 h

Desorbed at600 �C

7.3 0.15 29.74

Fig. 13. FTIR spectra of samples ball milled at 100 and 200 rpm afterisothermal desorption at 225 �C.

Fig. 14. FTIR spectra of samples ball milled at 100 rpm after isothermaldesorption at 225 �C.

5828 J.J. Hu et al. / Acta Materialia 59 (2011) 5821–5831

example, the N content of S200 rpm-10 h increased from27.47% to 28.55% after desorption at 225 �C and to30.00% after TPD at 600 �C. The volumetric H amountsobtained by Sieverts’ method were almost the same as theH available in these samples (7.3% vs 7.24% and 7.28%)measured by elemental analysis, suggesting the NH3 emis-sion was insignificant in the adequately milled samples.However, the N content decreased for the weakly ball-milled sample S100 rpm-2.5 h after release at 225 �C,implying that NH3 generation occurred already at225 �C. At 600 �C, the N content found in the samplewas 24.6%, much lower than that before dehydrogenation(27.1%), indicating a severe N loss at high temperature.

Based on the above results, we realized that the pre-dicted value of 8.2% H is not obtainable, even at high tem-peratures, unless high-purity MgH2 and LiNH2 are usedinstead of commercial products.

3.5. Desorption products

Akbarzadeh et al. [21] have systematically studied theLi–Mg–N–H systems using first-principles DFT calcula-tions. For the 1:1 mixture of LiNH2 and MgH2, theauthors proposed a multistep dehydrogenation processwith the following reaction sequence:

LiNH2 þMgH2 ! LiHþ 1=2MgðNH2Þ2 þ 1=2MgH2 ð8Þ! LiHþ 1=4MgðNH2Þ2 þ 1=4Mg3N2 þH2 ð9Þ! 1=2LiHþ 1=4Mg3N2 þ 1=4Li2MgðNHÞ2 þ 3=2H2

ð10Þ! LiMgNþ 2H2 ð11Þ

The first reaction to take place within the 1:1 mixturewas predicted to be the exothermic metathesis, whichagrees with our calorimetric results and simple thermody-namic calculation. However, different desorbed productsfrom the isothermal desorption at 225 �C were detected,depending on the ball-milling time and energy. The samplesS100 rpm-2.5 h and S100 rpm-2.5 h-TiCl3, and evenS100 rpm-48 h desorbed to an imide mixture of Li2NHand MgNH via reactions (3) and (6), respectively, as indi-cated by the vibration peaks at 3195 and 3165 cm�1 inthe FTIR spectra (Figs. 13 and 14). Phases of the uncon-sumed starting components LiNH2 and MgH2 were clearlyidentified in their XRD profiles (Fig. 15). This means thateven though the metathesis may take place first due to itsmore favorable thermodynamics and high driving force,other reactions which are thermodynamically less favorablecan overtake because of their faster kinetics. This is a typ-ical feature of solid-state reactions occurring in a heteroge-neous chemical environment.

Hence, the imide Li2Mg2(NH)3 was detected by XRD inthe desorbed products of briefly milled samples (Fig. 15),consistent with the observation by Osborn et al. [5]. Forthe samples with extended ball milling, the typical broadabsorbance at 3171 cm�1 points to the formation of the ter-nary imide Li2Mg(NH)2 via the well-known reaction

between LiH and Mg(NH2)2 (reaction (4)) [18]. Therefore,in practice the extent of metathesis induced by various ball-milling parameters determines the initial composition forthermal dehydrogenation and consequently the progressof dehydrogenation. Even after 5 wt.% H2 desorption, thediffraction peaks of MgH2 are still discernible in the sam-ples S100 rpm-96 h and S200 rpm-10 h (Figs. 16A and 17upper). These observations are clearly different to the reac-tion pathways predicted by DFT that indicate MgH2

would be consumed in the first dehydrogenation step withthe formation of Mg3N2. Although the dehydrogenationreaction between MgH2 and Mg(NH2)2 seems thermody-namically more favored than the reaction between LiHand Mg(NH2)2 [19,20], the reverse reaction order has infact been observed. Thus, the further the metathesis con-version has proceeded during the ball-milling step, themore hydrogen will be released via the kinetically faster

Fig. 15. XRD patterns of samples briefly ball milled at 100 rpm for 2.5 hafter isothermal desorption at 225 �C.

Fig. 16. XRD patterns of samples ball milled at 100 rpm after isothermaldesorption at 225 �C.

Fig. 17. XRD patterns samples ball milled at 100 rpm after desorption to600 �C.

J.J. Hu et al. / Acta Materialia 59 (2011) 5821–5831 5829

reactions (4) or (12) for the 1:1 system up to a maximum of4.1%.

LiNH2 þMgH2 ! LiHþ 1=2MgðNH2Þ2 þ 1=2MgH2

ð8Þ! 1=2Li2MgðNHÞ2 þ 1=2MgH2 þH2

ð12ÞAn amount of 3.7% H was measured for S100 rpm-96 h

and S200 rpm-10 h in TPD mode for the first inflexion(Fig. 12), which indicates that under the selected millingconditions the metathesis conversion was essentially com-pleted. Nevertheless, our finding that the desorption iso-therm at 225 �C went beyond 4.1% H hints at thethermodynamic feasibility of the dehydrogenation stepsseen in the DSC curves at higher temperatures, albeit withlow kinetics.

Although the XRD profiles of the corresponding sam-ples shown in Fig. 16 resemble the cubic structure ofLiMgN reported by Juza [22], an unambiguous assignmentis difficult, since the reflections from Mg3N2, LiMgN(JCPDS 72-1287), Li2NH, Li2Mg(NH)2 overlap at 2h posi-tions 30.7–31.1�, 51.8–52.0� and 61.0–61.8�, especially forthe amorphous samples with broad peaks. Recent investi-gations show that the fcc crystal structure (JCPDS 72-1287) described in Ref. [22] and identified by Lu et al.[4,8] has in fact the composition of Li1.12Mg0.88N0.96, whichis Li-rich. The crystal structure of LiMgN instead adoptedthe orthorhombic structure Pmna [23,24].

After desorbing at temperatures of up to 600 �C, thesamples without TiCl3 addition did present the orthorhom-bic structure Pmna as reported [23,24], whereas the TiCl3-doped samples rather resemble the Li-rich fcc structure(Fig. 18). It seems that the crystal structure of the desorbedproducts at 600 �C was affected by the addition of TiCl3.The presence of Mg3N2 in the desorbed products cannotbe excluded. At present, however, we are not able to

Fig. 18. XRD patterns of S200 rpm-10 h desorbed at 225 �C (upper) andto 600 �C (lower).

Fig. 21. First and second dehydrogenation isotherms at 225 �C for

5830 J.J. Hu et al. / Acta Materialia 59 (2011) 5821–5831

correlate the differences in crystal structure and dehydroge-nation resulting from the addition of TiCl3.

3.6. Rehydrogenation

Attempts to rehydrogenate the desorbed samples weremade on TiCl3-doped and undoped samples under differenttemperatures and H2 pressures. However, full rehydrogena-tion was not achieved. LiH and Mg(NH2)2 were indeeddetected after rehydrogenation at 202 �C and 140 bar(Fig. 19) by XRD, which nevertheless represents the mainfeatures of desorbed products including Mg3N2 as men-tioned above. Fig. 20 shows the DSC curves of S200 rpm-20 h from the as-prepared (top) and rehydrogenated sam-ples. The first endothermic peak relating to the dehydroge-nation reaction between LiH and Mg(NH2)2 seems to be atleast partially recoverable. Fig. 21 shows data of volumetric

Fig. 19. XRD pattern of S200 rpm-20 h-TiCl3 after rehydrogenation at202 �C and 140 bar.

Fig. 20. DSC curves of S200 r-10 h-TiCl3 of as-prepared (top) andrehydrogenated samples.

S200 rpm-10 h-TiCl3.

measurements of S200 rpm-10 h-TiCl3. Only 2 wt.% couldbe released after rehydrogenation at 163 �C and 108 barH2. These results are essentially the same as observed inby Liu [6] and Liang et al. [7], who reported a maximumof 2.3% rehydrogenation at 210 �C. They attributed theabsorption to the hydrogenation of 0.25 mol ternary imideLi2Mg(NH)2 in the desorbed products.

4. Conclusions

The mixture consisting of a 1:1 molar ratio of LiNH2

and MgH2 becomes a multicomponent system induced bymechanical ball milling. The metathesis reaction betweenLiNH2 and MgH2, resulting in the formation of Mg(NH2)2

and LiH, is believed to be the initial step for dehydrogena-tion to take place in this system. Furthermore, the millingtreatment is responsible for the extent of the metathesisreaction which determined the hydrogen desorption kinet-ics and pathways.

The addition of TiCl3 facilitated the metathesis step andthus dehydrogenation. However, at moderate temperature225 �C, only 5 wt.% H could be released. Even at 600 �C,a maximal amount of 7.3% H release was measured, belowthe predicted value of 8.1 wt.%.

A H-free desorption product of LiMgN with the ortho-rhombic Pmna structure was identified from the desorptionof non-doped samples, whereas the TiCl3-doped samplesdesorbed to products containing cubic Li-rich ternarynitride. Reversible hydrogen sorption was only partiallyrealized. Obviously, control over the desorption pathwayand an exact 1:1:1 stoichiometry of Li:Mg:N are crucialto making use of this binary mixture for hydrogen-storagepurposes.

Acknowledgements

Funding by the German–Chinese Sustainable Fuel Part-nership (GCSFP, Grant No. 03BV108A) and the EU Pro-

J.J. Hu et al. / Acta Materialia 59 (2011) 5821–5831 5831

ject NANOHy (Grant No. 210092) is gratefullyacknowledged.

References

[1] http://www.hydrogen.energy.gov/pdfs/review09/st_01_klebanoff.pdf.[2] http://www.hydrogen.energy.gov/pdfs/review09/st_04_fang.pdf.[3] Alapati SV, Johnson JK, Sholl DS. J Phys Chem B 2006;110:8769.[4] Lu J, Fang ZZ, Choi YJ, Sohn HY. J Phys Chem C 2007;111:12129.[5] Osborn W, Markmaitree T, Shaw LL. J Power Sources 2007;172:376.[6] Liu Y, Zhong K, Gao M, Wang J, Pan H, Wang Q. Chem Mater

2008;20:3521.[7] Liang C, Liu Y, Luo K, Li B, Gao M, Pan He, et al. Chem Eur J

2010;16:693.[8] Lu J, Choi YJ, Fang ZZ, Sohn HY. J Power Sources 2010;195:1992.[9] Luo W. J Alloys Compd 2004;381:284.

[10] Chen P, Xiong Z, Luo J, Lin J, Tan KL. Nature 2002;420:302.[11] Luo W, Ronnebro E. J Alloys Compd 2005;404:392.[12] Hu JJ, Fichtner M, Chen P. Chem Mater 2008;20:7089.

[13] Aylward G, Findlay T. Thermodynamic database: SI chemical data.4th ed. Singapore: John Wiley and Sons; 1998.

[14] Rijssenbeek J, Gao Y, Hanson J, Huang Q, Jones C, Toby B. J AlloysCompd 2008;454:233.

[15] Choi YJ, Lu J, Sohn HY, Fang ZZ. J Power Sources 2008;180:491.[16] Paskevicius M, Sheppard DA, Buckley CE. J Am Chem Soc

2010;132:5077.[17] Juza R, Opp K. Z Anorg Allg Chem 1951;266:325.[18] Xiong Z, Hu JJ, Wu G. J Alloys Compd 2005;398:235.[19] Hu JJ, Xiong Z, Wu G, Chen P. J Power Sources 2006;159:120.[20] Hu JJ, Wu G, Liu Y, Xiong Z, Chen P. J Phys Chem B

2006;110:14688.[21] Akbarzadeh AR, Ozolins V, Wolverton C. Adv Mater 2007;19:3233.[22] Juza R. Naturwissenschaften 1946;33:121.[23] Yamane H, Okabe TH, Ishiyama O, Waseda Y, Shimada M. J Alloys

Compd 2001;319:124.[24] Bailey AS, Hubberstey P, Hughes RW, Ritter C, Gregory DH. Chem

Mater 2010;22:3174.