IOP PUBLISHING NANOTECHNOLOGY

Nanotechnology 19 (2008) 465706 (5pp) doi:10.1088/0957-4484/19/46/465706

The Pd catalyst effect on low temperaturehydrogen desorption from hydridedultrathin Mg nanobladesF Tang1,2, T Parker2, H-F Li, G-C Wang and T-M Lu

Department of Physics, Applied Physics and Astronomy, Rensselaer Polytechnic Institute,110 8th Street, Troy, NY 12180-3590, USA

E-mail: ftang7@exchange.asu.edu

Received 11 May 2008, in final form 10 September 2008Published 22 October 2008Online at stacks.iop.org/Nano/19/465706

AbstractWe present hydrogenation/de-hydrogenation properties of ultrathin Mg nanoblades coated withPd as a catalyst. The hydrided Pd/Mg/Pd and Pd/Mg nanoblades were highly curved asobserved by ex situ scanning electron microscopy. An in situ thermal desorption spectroscopy(TDS) study showed that the hydrided Pd/Mg/Pd nanostructure has a low hydrogen desorptiontemperature, at ∼365 K. Through a combinational microstructure and TDS analysis of hydridedPd/Mg/Pd nanoblades as well as hydrided Pd/Mg/Pd nanoblades covered with an additionalultrathin Mg layer, we found that the effect of Pd catalyst on reducing the hydrogen desorptiontemperature is significantly stronger than the conventionally proposed mechanisms due to grainsize and strain.

(Some figures in this article are in colour only in the electronic version)

1. Introduction

Magnesium is a very attractive material for hydrogen storagebecause it is abundant and has a low density (1.738 g cm−3).Theoretically, magnesium hydride (MgH2) contains 7.6 wt%of hydrogen, which is above 6 wt% required by DOE hydrogenplans. However, due to the ionic bond nature between Mg andH atoms, MgH2 is thermodynamically stable [1]. Additionally,the hydrogenation/de-hydrogenation of Mg is also kineticallyhindered by a number of barriers. For example, the hydrogenmolecules do not readily dissociate on the Mg surface with anenergy barrier of ∼1.1 eV [2, 3]. Hydrogen atoms also havean extremely slow diffusion in magnesium hydride. All thesefactors make the hydrogen kinetics of magnesium sluggishunless the material is heated to ∼573–673 K.

Various strategies have been explored to enhance thehydrogen kinetics of magnesium hydride. One methodis to fabricate the Mg particles to nanometers scale size.For example, using mechanical ball milling, Zaluska et al[4, 5] have created nanocrystalline Mg and Mg2Ni, whichshowed one order of magnitude faster kinetics of hydrogen

1 Author to whom any correspondence should be addressed.2 The first two authors contributed equally to this work.

absorption and desorption than their corresponding bulkmaterials. However, the problems, such as contamination andoxidation during the mechanical ball milling, remain to besolved [6]. In addition, the bulk Mg material can hardly beground into particles with a diameter less than 100 nm. Veryrecently, Mg nanowires with a diameter down to ∼30 nm havebeen fabricated by a vapor-transport approach [7]. The additionof a catalyst is another method used extensively to overcomethe inherent limitations of pure Mg [8–10]. Since catalystsbasically work at the surface of materials, their functionalityis very sensitive to contamination and oxidation on surfaces.Pd Coated Mg thin films have been demonstrated to havethe most impressive hydrogen kinetics with a low desorptiontemperature peak of ∼373 K [11–14]. The enhanced catalystfunctionality is essentially attributed to the film preparation ina high vacuum condition, which gives minimum contaminationand oxidation on surfaces. However, the continuous filmstructure limits the catalytic effect to the top ∼200 nm ofthe film leaving the rest of the Mg film without this catalyticadvantage [13].

In this paper, we reported hydrogenation/de-hydrogenationproperties of ultrathin Mg nanoblades (∼15 to ∼30 nm thick)coated with Pd as a catalyst. This unique composite nanostruc-

0957-4484/08/465706+05$30.00 © 2008 IOP Publishing Ltd Printed in the UK1

Nanotechnology 19 (2008) 465706 F Tang et al

ture has shown a low hydrogen desorption temperature peak at∼365 K. Through the analyses of hydrided nanoblades by exsitu scanning electron microscopy (SEM), transmission elec-tron microscopy (TEM) and in situ thermal desorption spec-troscopy (TDS), we have revealed that the Pd catalyst plays amore important role in reducing the hydrogen desorption tem-perature than other possible factors, such as grain size andstrain [11, 14].

2. Experiments

The Pd coated nanoblades were prepared in an ultrahighvacuum (UHV) chamber equipped with Mg and Pd thermalevaporation sources. We first grew Mg nanoblades on theSi substrate by oblique angle deposition (OAD). Duringdeposition the Mg vapor flux was incident on the substrate at75◦ with respect to the substrate normal [15]. The depositednanoblades are nearly vertically standing with an angle slightlytilted away from the flux [16]. In order to coat Pd over theentire length Mg nanoblades, the substrate was first rotatedabout the substrate normal, so that the vertical Mg nanobladeswere approximately aligned with the Pd flux direction. Thenduring the Pd evaporation we either coated one side of thenanoblades with the substrate stationary or swung the substratepolarly in an angular range of ∼15◦ to achieve the coatings onboth sides of the nanoblades. During the growth of Mg andPd the pressure increased from the base pressure of ∼3.0 ×10−9 Torr to ∼2.0 to 3.0 × 10−8 Torr. The evaporation rateof Mg was maintained between ∼9.0 and ∼12.0 nm min−1 asindicated by a quartz crystal monitor (QCM). The Pd growthrate was ∼0.12 nm min−1 and a total of ∼4 to 6.5 nm Pd wasdeposited based on the reading from the QCM. After the Pdcoating, the Mg nanoblades were transported in situ from theevaporation chamber to the reaction chamber connected by anin-line valve. The composite nanoblades were hydrogenated ata pressure of 1 bar with substrate held at ∼333 K for ∼15 h.The TDS was used to investigate the de-hydrogenation kineticsunder a base pressure ∼1.5 × 10−8 Torr after venting anddegassing the reaction chamber. The temperature rising ratefor TDS was ∼3 K min−1. A residue gas analyzer (RGA)was used to measure the amount of desorbed hydrogen gas bymonitoring the H2 partial pressure as a function of time (ortemperature through heating rate).

3. Results and data analyses

3.1. SEM analyses

Figure 1(a) shows the ex situ SEM cross-section image of theas deposited Mg nanoblade film with one side of nanobladecoated with Pd (named Pd/Mg nanoblades). From the image,we can see that the as deposited nanoblades were slightlycurved. After the hydrogenation in the reaction chamber, thePd/Mg nanoblades became highly curved and bent obviouslytowards the incident vapor fluxes, as shown in figure 1(b).It is known that there is a ∼30% volume expansion duringthe transformation from Mg to MgH2. This volume changecan yield a considerable stress [14, 17, 18] and cause the

deformation of nanostructures. Since only one side of eachnanoblade was coated with Pd, we may intuitively think the Pdcoated side was more hydrided and extended more. However, ifthis picture was correct, the nanoblades should bend to the sidewithout Pd coating, specifically these should bend away fromthe fluxes, which is opposite to our experimental observations.It was also reported [13] that with a Pd coating the Mg thinfilms can be hydrided up to a depth of ∼200 nm beneaththe surface. Therefore, the entire thickness of nanoblades(∼15 to ∼30 nm) around the Pd coating very likely hasbeen fully hydrided. Another possible mechanism for thecurvature formation of nanoblades is the decomposition ofthe hydrided nanoblades. From previous studies [11, 12],it was shown that the MgD2 film with Pd coating partiallywas decomposed into Mg in air. The MgD2 has an identicalchemical property as MgH2. In this case the Mg hydride nearthe Pd coating side will be decomposed, which results in thebending of nanoblades towards the flux. This picture is moreconsistent with our experimental observations. The furtherdecomposition of MgH2 may be prevented by the Mg oxidelayer formed between Pd and MgH2 in air [19]. Assumingthe formation of a bilayer with an equal thickness of Mg andMgH2, then the strain produced due to the decomposition ofMgH2 can be calculated from 2t/(3R), where R is the radiusof the curvature [20]. By fitting the curliest nanoblade, asshown in figure 1(c) we can obtain the radius of the curvature,R ∼ 200 nm. Then the calculated strain required to producethis curvature is equal to 2t/(3R) = 6.6%, by using an averaget of ∼20 nm. This value of required strain is smaller thanthe value that can be theoretically produced by MgH2 and Mgbilayer, which is ∼8.37%. This theoretical strain is calculatedfrom (1 − (1/1.3)1/3) = 8.37% assuming a ∼30% isotropicvolume change during the decomposition.

In addition to the circularly curved nanoblades, manyhelical structures also exist, shown in the SEM top viewimage of figure 1(c). This helical structure is due to thenon-uniform coating of Pd on the particular nanoblades. Forexample, if the non-uniform coating produces a strain alongthe diagonal direction instead of the edge of the nanoblades,then the nanoblades will curve into a helical shape to releasethe generated strain [20].

Figure 1(d) shows an ex situ SEM image of hydridedMg nanoblades with Pd coating on both sides (Pd/Mg/Pdnanoblades). From the image we can see that the nanobladesare less curved and there is no obviously preferred bendingdirection any more. Since the Pd coating along the nanobladelength is not uniform, then at a particular height of a nanobladethe coating of Pd on both sides of a nanoblade can be verydifferent. However, this difference in coating varies along thenanoblade length, resulting in the disappearance of a preferredbending direction. In our later analysis, we will concentrate onthe nanoblades with Pd coating on both sides, which is moreresistive to the oxidation in air.

3.2. TEM images

The microstructure of the air exposed hydrided nanobladeswith Pd coating on both sides was further examined ex situ

2

Nanotechnology 19 (2008) 465706 F Tang et al

Flux

(c)

1 m

Mg Pd

1 m

Mg Pd(b)

R

1 μ

μ

μ

m

(a) Mg Pd

(d)

Figure 1. Ex situ (a) SEM cross-sectional view of the as depositedMg nanoblades with one side coated with Pd (Pd/Mg nanoblades);(b) SEM cross-sectional view of the hydrided Pd/Mg nanoblades.The R is the radius of the curved hydrided Pd/Mg nanoblades. Thearrows indicate the incident flux directions for Mg and Pd. (c) SEMtop view of the helical structures formed in hydrided Pd/Mgnanoblades film. (d) SEM cross-sectional view of hydrided Mgnanoblades with both sides coated with Pd (Pd/Mg/Pd nanoblades).The arrows near label Pd indicate the incident Pd flux directionrelative to the substrate normal when the substrate was swung about15◦. The double headed arrow represents the polar angular range ofthe sample swing.

using TEM. Figure 2(a) shows the bright field imageof hydrided nanoblades. Using the electron energy lossspectroscopy (EELS) mapping mode in TEM, we investigatedthe Pd element distribution in the nanoblades. The inset offigure 2(a) is a typical Pd distribution map obtained by usingthe Pd M4,5 peak (335 eV). The bright areas in the map containa higher Pd concentration than the dark areas. The mapindicates a non-continuous coating of Pd on the nanoblades.Since the thickness of the Pd coating on the nanoblades was onthe order of a few nanometers, the growth may still be in theisland nucleation stage. Moreover, the Pd flux was obliquelyincident on the side faces of the nanoblades thereby enhancingthe island formation due to shadowing effect [21–23]. Weestimated the average Pd coverage from the multiple Pd EELSimages of the nanoblades. The analysis software used for

3.29 0.07 ≈

2.29 0.06 ≈

MgH2 (110)

MgH2 (200)/Pd (111)

3.29± 0.07 Å

2.29± 0.06 Å

MgH2 (110)

Pd (111)/ MgH2 (200)

0.5 μm

Pd (111)

MgH2 (110)

MgH2 (200)

Mg (1011) -

Mg (1010) -

MgO (200)

(b) (c) MgH2 (110)

MgH2 (200)

Mg (1011) -

Mg (1010) -

MgO (200)

Pd (111)

(a)

(d)

b c

Figure 2. (a) Ex situ TEM bright field image of Pd/Mg/Pdnanoblades after hydrogenation. The inset of (a) is an EELS mappingof Pd. The bright areas in the map represent higher Pd coverage thanthat in the dark areas; (b) and (c) are SAED patterns from differentareas indicated by the circles on the hydrided nanoblade; (d) a latticefringe image near one edge of the hydrided nanoblade.

calculating the coverage is UTHSCSA ImageTool [24]. Inthe calculation from one Pd EELS image we considered thatthe dark area of the image has no Pd coverage and the brightand white area with EELS intensity greater than zero has Pdcoverage. The estimated coverage was calculated by taking theratio between the total area of all the bright and white regionsand the total area of the image. The coverage of Pd coatingvaried along the length of the nanoblade, which yields a largestandard error of 23% in the estimated value of 60 ± 23% frommultiple Pd EELS images.

This Pd coating is very important to the hydrogenationprocess as the hydrogen molecules were dissociated by Pd to

3

Nanotechnology 19 (2008) 465706 F Tang et al

provide atomic hydrogen [19, 25]. As we will show later, Pdalso plays a critical role in assisting the hydrogen desorption inthe de-hydrogenation process.

Figures 2(b) and (c) are selected area electron diffraction(SAED) patterns taken from the circled regions on the hydridednanoblade. The patterns consist of diffraction rings and spots.By measuring the corresponding reciprocal lattice distancesin the diffraction pattern, we identified the crystalline phasesand indexed the diffraction pattern as shown in figures 2(b)and (c). The obvious diffraction rings of MgH2(110) andMgH2(200) indicate a significant hydride in the air exposednanoblade. We can also see the Mg phase in this air exposedsample as represented by diffraction spots. However thesediffraction spots scattered and did not form a single crystallinepattern as in the pure Mg nanoblades [15]. This showsthat the remaining Mg crystallites have been reoriented [15].Figure 2(d) shows the high resolution transmission electronmicroscopy (HRTEM) image taken from the region near theedge of a nanoblade. The lattice fringes in the image mainlycorrespond to two interplanar distances from the measurement.One is 3.29±0.07 A, which can only match with the interplanardistance of (110) planes of MgH2 at 3.19 A. Some of thesegrains are outlined by white dashed curves. The size of theMgH2(110) grains is ∼(5.3 ± 2.0) nm. The other one hasan interplanar distance of 2.29 ± 0.06 A, which is close tothe interplanar distances of MgH2(200) and Pd(111) which are2.26 A and 2.24 A, respectively. These grains are outlined bywhite solid curves. In addition, we can see from figure 2(d) thata thin oxide layer ∼2 to ∼5 nm thick in the boundary regionof the nanoblades exists. This thin oxide layer has a nearlyamorphous structure, which is consistent with the very weakoxide diffraction rings in figures 2(b) and (c).

3.3. TDS measurement

The de-hydrogenation property of hydrided Pd/Mg/Pd nanoco-mposite structure was in situ characterized by TDS. Figure 3shows the TDS spectrum of hydrided nanoblades. Two wellseparated hydrogen desorption peaks were observed in thespectrum. This suggests the existence of two different kineticprocesses of hydrogen desorption. The major desorption peakoccurs at ∼365 K while the other peak lies at a significantlyhigher temperature of ∼513 K. Repeated experiments onsamples prepared under similar conditions showed that thesetwo peaks can vary from ∼365 to 393 K and ∼503 to 560 K,respectively. Researchers have found the low temperaturedesorption peaks at the Pd/Mg thin film structure, which rangesfrom ∼388 to 483 K [11, 13]. However, the mechanismcausing this desorption peak is still being debated. Differentfactors, such as Pd catalyst [9], crystal size and strain havebeen proposed [11, 13]. For a nanostructure, the strain canbe mostly released through the free surfaces. Schimmelet al also demonstrated that the strain may not be the majorfactor for the enhanced hydrogen adsorption and desorptionkinetics [26]. Another factor like grain size can significantlylower hydrogen desorption temperature only when the size ofthe MgH2 clusters is less than 1 nm [27]. Such a size is muchsmaller than our experimentally observed MgH2 grain size of∼(5.3 ± 2.0) nm.

300 350 400 450 500 550 6000

4

8

12

16

20

24

28

3250 100 150 200 250 300

Temperature (oC)

Mg/hydrided nanoblades

Hydrided nanoblades

Temperature (K)

H2 p

artia

l pre

ssur

e (1

0-8 T

orr)

Figure 3. In situ TDS spectra of hydrided Pd/Mg/Pd nanoblades(blue curve) and Mg coated hydrided Pd/Mg/Pd nanoblades(red curve).

4. Discussion

Previous studies have illustrated the critical role of Pd catalystin enhancing the hydrogenation kinetics of Mg. However,the effect of Pd on the de-hydrogenation process has beenless studied [9]. To test this effect, we coated another thinlayer of Mg on the hydrided Pd/Mg/Pd nanoblades. Our wayof coating this Mg layer is basically similar to that of Pdcoating, except with a larger substrate swing angular range(∼25◦) and a much thicker Mg layer of ∼25 nm, in contrastwith ∼6.5 nm thick Pd coating. The catalyst effect of Pd onthe de-hydrogenation process is expected to be significantlyreduced due to this additional Mg coverage. In the mean time,the grain size of MgH2 and strain remain unchanged in theseMg covered Pd/Mg/Pd nanoblades. From the TDS spectrumof the Mg coated hydrided Pd/Mg/Pd nanoblades shown infigure 3, we can clearly see that the first hydrogen desorptionpeak has been shifted to a significantly higher temperature of∼503 K. This increased desorption temperature should be dueto the loss of Pd catalyst effect. Although the detailed atomicprocess of catalysis is not clear, we believe that Pd may helpthe recombination process of H atoms into hydrogen moleculeson the surface [3, 11]. Through the simulation prediction [3],Vegge has shown that the recombination barrier for a hydrogenmolecule is the limiting factor for hydrogen desorption frompure Mg surface. The coated Mg may require H atoms todiffuse through an extra layer of Mg with a thickness upto ∼12.5 nm (half of ∼25 nm). However, this diffusionthrough a distance of tens of nanometer will take less than amicrosecond [11]. Therefore, it is not an impeding factor to thehydrogen desorption process. We also observed another peakat ∼560 K, which is close to the desorption peak of pure MgH2

at ∼573 K [1]. This indicates that the Pd layer has almost fullylost its catalyst effect at this temperature.

The diminishing of Pd catalyst effect may also accountfor the higher temperature desorption peak observed fromthe hydrided Pd/Mg/Pd nanoblades. In the low temperature

4

Nanotechnology 19 (2008) 465706 F Tang et al

desorption peak of ∼365 K, the Pd can serve as a catalyst.However, after certain H desorption, the pure Mg will beproduced around Pd. From the Pd mapping, we know thatthe Pd coating is not continuous, the deposited pure Mg willmigrate and cover Pd clusters. This migration has been widelyfound [19, 28, 29], Krozer et al [19] has shown that at atemperature of ∼353 K, Mg can migrate through 23 nm thickPd coating and cover the top of Pd surface. Pd and Mg canalso form alloys in the interface region at a temperature as lowas 323 K [30, 31]. Once the Pd surface was fully modified toform the alloy through the migration then hydrogen would haveto be released at a much higher temperature. For air exposedPd/Mg/Pd nanoblades, the obviously alloying between Pd andMg starts at ∼480 K, which is significantly delayed. This delaycould be due to the oxide layer formed between Pd and Mgduring the air exposure [32].

5. Conclusion

In summary, we have observed the air exposed hydridedPd/Mg/Pd and Pd/Mg nanoblades were highly curved.Through the analyses of hydrided Pd/Mg/Pd nanoblades andMg covered hydrided Pd/Mg/Pd nanoblades by ex situ SEM,TEM and in situ TDS, we have revealed the importanceof Pd catalyst on the hydrogen desorption process at lowtemperature. The understanding of the effect of Pd catalyston the hydrogen desorption process could help designing ofpromising hydrogen storage nanoscale metal hydrides with alow hydrogen desorption temperature.

Acknowledgments

FT was supported by NSF award 0506738. TP was supportedthe DOE (education) GAANN P200A030054. We thankProfessor Sheng-Bai Zhang for his invaluable discussions onthe results of hydrogenation/de-hydrogenation measurements.

References

[1] Tsuda M, Dino W A, Kasai H, Nakanishi H andAikawa H 2006 Thin Solid Films 509 157

[2] Grochala W and Edwards P P 2004 Chem. Rev. 104 1283[3] Vegge T 2004 Phys. Rev. B 70 035412[4] Zaluska A, Zaluski L and Stro-Olsen J O 1999 J. Alloys

Compounds 288 217[5] Zaluska A, Zaluski L and Stro-Olsen J O 2001 Appl. Phys. A

72 157

[6] Akyıldız H, Ozenbas M and Ozturk T 2005 Hydrogen storagein magnesium based thin films Int. Hydrogen Energy Congr.(Istanbul) pp 1–9

[7] Li W-Y, Li C-S, Ma H and Chen J 2007 J. Am. Chem. Soc.129 6710

[8] Zlotea C, Lub J and Andersson Y 2006 J. Alloys Compounds426 357

[9] Berlouis L E A, Honnor P, Hall P J, Morris S and Dodd S B2006 J. Mater. Sci. 41 6403

[10] Domenech-Ferrer R, Sridharan M G, Garcia G, Pi F andRodrıguez-Viejo J 2007 J. Power Sources 169 117

[11] Checchetto R, Bazzanella N, Miotello A, Brusa R S,Zecca A and Mengucci A 2004 J. Appl. Phys. 95 1989

[12] Checchetto R, Brusa R S, Bazzanella N, Karwasz G P,Spagolla M, Miotello A, Mengucci P and Cristoforo A D2004 Thin Solid Films 469/470 350

[13] Higuchi K, Kajioka H, Toiyama K, Fujii H, Orimo S andKikuchi Y 1999 J. Alloys Compounds 293–295 484

[14] Higuchi K, Yamamoto K, Kajioka H, Toiyama K, Honda M,Orimo S and Fujii H 2002 J. Alloys Compounds330–332 526

[15] Tang F, Parker T, Li H-F, Wang G-C and Lu T-M 2007J. Nanosci. Nanotechnol. 7 3239

[16] Tang F, Wang G-C and Lu T-M 2007 J. Appl. Phys.102 014306

[17] Leon A, Knystautas E J, Huot J and Schulz R 2002 J. AlloysCompounds 345 158

[18] Leon A, Knystautas E J, Huot J and Schulz R 2006 Thin SolidFilms 496 683

[19] Krozer A and Kasemo B 1990 J. Less-Common Met. 160 323[20] Huang M-H, Boone C, Roberts M, Savage D E, Lagally M G,

Shaji N, Qin H, Blick R, Nairn J A and Liu F 2005 Adv.Mater. 17 2860

[21] Tait R N, Smy T and Brett M J 1993 Thin Solid Films 226 196[22] Robbie K and Brett M J 1997 J. Vac. Sci. Technol. A 15 1460[23] Dirks A G and Leamy H J 1997 Thin Solid Films 47 219[24] http://ddsdx.uthscsa.edu/dig/itdesc.html[25] Du A J, Smith S C, Yao X D and Lu G Q 2007 J. Am. Chem.

Soc. 129 10201[26] Schimmel H G, Johnson M R, Kearley G J,

Ramirez-Cuesta A J, Huot J and Mulder F M 2005 J. AlloysCompounds 393 1

[27] Wagemans R W P, Lenthe J H v, Jongh P E d, Dillen A J v andJong K P d 2005 J. Am. Chem. Soc. 127 16675

[28] Kelekar R, Giffard H, Kelly S T and Clemens B M 2007J. Appl. Phys. 101 114311

[29] Yoshimura K, Yamada Y and Okada M 2004 Surf. Sci. 566 751[30] Krozer A, Fischer A and Schlapbach L 1996 Phys. Rev. B

53 13808[31] Slack J L, Locke J C W, Song S-W, Ona J and Richardson T J

2006 Sol. Energy Mater. Sol. Cells 90 485[32] Tang F, Yuan W, Lu T-M and Wang G-C 2008 J. Appl. Phys.

104 033534

5