decomposition of tih2 studied in situ by synchrotron x-ray and neutron diffraction

$: Decomposition of TiH2 studied in situ by synchrotron X-ray and neutron diffraction$
Available online at www.sciencedirect.com

www.elsevier.com/locate/actamat

Acta Materialia 59 (2011) 6318–6330

Decomposition of TiH2 studied in situ by synchrotron X-rayand neutron diffraction

C. Jimenez a,b,⇑, F. Garcia-Moreno a,b, B. Pfretzschner a,b, M. Klaus b, M. Wollgarten b,I. Zizak b, G. Schumacher b, M. Tovar b, J. Banhart a,b

a Technische Universitat Berlin, Hardenbergstr. 36, 10623 Berlin, Germanyb Helmholtz-Zentrum Berlin fur Materialien und Energie, Hahn-Meitner-Platz 1, 14109 Berlin, Germany

Received 26 April 2011; received in revised form 23 June 2011; accepted 24 June 2011Available online 23 July 2011

Abstract

Decomposition during constant heating of as-received and pre-oxidized titanium hydride (TiH2) powder under flowing and stationaryAr was studied in situ by synchrotron X-ray and neutron diffraction. Diffraction data were compared to corresponding thermoanalysistraces in order to relate peaks of H2 release with individual phase transformations. We found that every peak of H2 release is correlatedwith regimes of lattice parameter contraction. Core–shell models to describe the decomposition processes measured were found, and fur-ther verified by electron microscopy. The models show that a shell of the hexagonal close-packed hydrogen solid solution (a) around eachparticle core controls its dehydrogenation for as-received TiH2, while a TiO2 rutile shell controls the outgassing from the pre-oxidizedpowder.� 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: TiH2 decomposition; In situ synchrotron XRD; Blowing agent; Core–shell model

1. Introduction

TiH2 powder is the most widely used blowing agent formaking metal foams either through powder metallurgical(PM) or melting routes [1]. It is also used for producingporous Ti for biomedical applications [2] and as ahydrogen-storage medium [3]. Metal foams made by thePM route have more regular pore sizes if TiH2 starts torelease H2 during melting of the alloy [4,5]. Dependingon the alloy used, TiH2 is therefore commonly used eitherin the as-received condition or after pre-treatments.Prominent H2 release accompanied by mass loss from as-received TiH2 starts at about 400 �C and peaks at varioustemperatures ranges. After performing an oxidation pre-

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

doi:10.1016/j.actamat.2011.06.042

⇑ Corresponding author at: Technische Universitat Berlin, Harden-bergstr. 36, 10623 Berlin, Germany.

E-mail address: [email protected] (C. Jimenez).

treatment, the onset temperature of TiH2 can be shiftedup to between 500 and 550 �C, and the H2 release curveis modified to exhibit a single peak [5–9]. Core-reactingmodels were applied to describe the oxidation process[9,10] but the decomposition processes of both as-receivedand pre-oxidized TiH2 powders are not yet fully under-stood. It is still unclear how the peaks of H2 release arerelated to the phase transformations of TiH2.

The combination of thermoanalysis with in situ diffrac-tion is appropriate for such investigations. Liu et al. [3]and Sandim et al. [11] studied the dehydrogenation ofTiH2 powder in this way. They performed detailed ther-moanalysis but achieved a limited temperature resolutionof 50 K for in situ angle-dispersive X-ray diffraction (AD-XRD) of Cu Ka radiation from X-ray tubes. This resultedin misinterpretations of their diffraction data that ledthem to report incorrectly the phase transformationsequences. An additional drawback common to bothstudies was the use of different atmospheric conditions:

rights reserved.

http://dx.doi.org/10.1016/j.actamat.2011.06.042

mailto:[email protected]


C. Jimenez et al. / Acta Materialia 59 (2011) 6318–6330 6319

inert gas flow for thermoanalysis and vacuum for in situXRD. This difference prevented a direct comparison ofthe results from the two methods because decompositionis significantly faster under vacuum than in flowing inertgas [12].

Energy-dispersive X-ray diffraction (ED-XRD) ofwhite synchrotron radiation in transmission geometrywas used as the main method in this study because itallows for much higher acquisition rates than AD-XRDfrom X-ray tubes due to higher flux and simultaneousacquisition of diffractograms. This allowed for a tempera-ture resolution of 2.9 K and a clearer insight into thephase transformation sequence of as-received TiH2. Inaddition, disputed aspects of the sequence were clarifiedboth with AD-XRD of monochromatic synchrotron radi-ation and in situ neutron diffraction. The rationale ofusing three different diffractometers is to vary the probingdepth of the radiation and the possibilities of adjustingthe atmospheric conditions around the decomposinghydride. This study reports for the first time the phasetransformation sequence of pre-oxidized TiH2, and a cor-relation between phase transformation sequences andpeaks of H2 release. The spatial distribution of phasesinvestigated by electron microscopy validated experimen-tally core–shell models for the decomposition of as-received as well as pre-oxidized TiH2 powders under con-ditions of constant heating.

2. Experimental

2.1. Materials: as-received and pre-oxidized TiH2 powder

We used TiH2 powder supplied by Chemetall GmbH,Germany (PS grade, purity 98.8%, nominally <36 lm,D50 = 6 lm). We prepared pre-oxidized TiH2 powder byheating 3 g of as-received powder in an alumina crucibleinside an open horizontal tube furnace pre-heated to480 �C. After 180 min, the crucible was withdrawn fromthe furnace and cooled down to ambient temperature.

2.2. Thermoanalysis and oxygen content analysis

Thermogravimetry and mass spectroscopy (TG-MS)were carried out simultaneously in a Netzsch STA 409 C,which is a thermobalance coupled via a skimmer to a quad-rupole mass spectrometer. The precision of the thermobal-ance is 10�4 g. A blank measurement was performed on anempty alumina crucible heating at 10 K min�1 from 35 to1100 �C under flowing Ar gas. Before heating, the ther-moanalyzer was twice evacuated and back-filled with Ar.After the blank measurement, 200 mg of powder werepoured into the crucible and TG-MS analysis was doneapplying the same temperature profile and atmosphereused for the blank. Oxygen contents were determined bycarrier gas hot extraction in a Horiba EMGA 620 WCnitrogen/oxygen analyzer on 5 mg powder [13].

2.3. Ex situ AD-XRD

Powder AD-XRD of Cu Ka radiation was performed exsitu on as-received and pre-oxidized TiH2 samples in a Bru-ker-AXS D8 Advance two-circle diffractometer. These dif-fractograms were used for phase identification with thehelp of the International Center for Diffraction Data(ICDD-PDF) database. Lattice parameters were deter-mined by Rietveld refinements using the software FullProf[14].

2.4. In situ diffraction experiments

2.4.1. ED-XRDThe phase-transformation sequences were followed

in situ by ED-XRD at the EDDI experimental station atthe BESSY II synchrotron light source of the HelmholtzCentre Berlin. The powder sample was illuminated by awhite beam of X-rays and the energy of diffracted photonswas measured in transmission geometry at 2h = 8� by amultichannel analyzing detector. Using this method, thewhole diffracted spectrum is available at all times. Furtherdetails of the EDDI beamline are given in Refs. [15,16]. Forthis study, an Anton Paar DHS 1100 furnace equippedwith an X-ray-transparent graphite dome was attached tothe goniometer. Powder samples were poured into a thincylindrical alumina crucible of 5 mm inner diameter,through the wall of which a hole had been laser-drilled.After tightening the crucible to the heating plate, a thermo-couple was slid through the hole to measure the sampletemperature, TS. A second thermocouple measured thetemperature of the heating plate, TH. The graphite domeclosed the furnace chamber which was evacuated for5 min using a mechanical pump and backfilled with Ar.An Ar gas pressure of 1.2 bar at the inlet induced Ar flowthrough the chamber. The temperature was increased from30 to 950 �C at 10 K min�1, after which the heater wasturned off and cooling took place. Additionally, as-receivedpowder samples were heated from 30 to 950 �C at5 K min�1 in flowing Ar and at 10 K min�1 but in a sta-tionary Ar atmosphere. The latter was realized by closingthe exhaust and gas inlet before heating. Data acquisitionand temperature cycle were computer-controlled by thesoftware package Spec [17]. One spectrum was acquiredevery 17.4 s; thus the temperature resolution for heatingat 10 K min�1 was 2.9 K.

2.4.2. AD-XRD

As AD-XRD in reflection allows surface-specific effectsto be assessed, the phase transformation sequence of as-received TiH2 was followed in situ also in reflection geom-etry at the KMC-2 experimental station at BESSY II of theHelmholtz Centre Berlin. A 2-D detector covered a 2h win-dow of 15� and acquired one diffractogram every 11 s,yielding a temperature resolution below 2 K. The powderwas cold-pressed to tablets of 6 mm diameter and 1 mmthickness that were fastened to a resistive heater. A photon

6320 C. Jimenez et al. / Acta Materialia 59 (2011) 6318–6330

energy of 8 keV provided about 11 lm penetration depththrough Ti. A cavity drilled at the sample surface accom-modated a thermocouple beside the beam path. A domewith X-ray-transparent Kapton windows sealed a chamberduring isochronal heating at 10 K min�1 under Ar flow.

2.4.3. Neutron diffraction

Neutron powder diffraction in transmission mode wasused to study in situ and in the bulk the phase transforma-tion of TiD2 at the diffractometer E6 of the Helmholtz Cen-tre Berlin [18]. Neutrons of 0.24 nm wavelength and anarea detector covering a 2h window of 15.5�, fixed and cen-tred at 2h = 62�, were used to perform in situ diffraction atvarying temperatures at 10 K min�1 between 348 and897 �C. The temperature resolution was 20 K. TiD2 insteadof TiH2 was used to reduce incoherent scattering fromhydrogen, thus reducing background noise. Three gramsof as-received TiD2 powder (made by GfE, Nuremberg)was poured into a quartz tube connected to a gas-tightreceptacle of 20 l capacity to limit the pressure increasedue to outgassing. A valve kept the gas pressure below1.2 bar. After establishing a stationary Ar atmosphere,the quartz tube was positioned inside a furnace with openpaths for incoming and diffracted neutrons. At 348 and897 �C full 2h scans were also performed.

2.5. Electron microscopy

Scanning electron microscopy (SEM) was carried outusing a Zeiss SUPRA VP operated at 10 kV acceleratingvoltage. Transmission electron microscopy (TEM) was per-formed in a Zeiss LIBRA 200 FE operated at 200 kV accel-erating voltage. Bright-field images and electron diffractionpatterns were obtained using a CCD camera. An in-columnenergy filter was employed to obtain zero-loss filteredbright-field (ZLFBF) images and selected-area electron dif-fraction (SAED) patterns. Throughout the paper, SAEDpatterns are shown with an inverted greyscale to improvethe visibility of faint diffraction features. The energy filterwas also used for elemental mappings by acquisition ofjump ratio images at the oxygen K-edge. For sample prep-aration, the powders were mixed with an epoxy resin con-taining carbon nanoparticles. After curing, slices were cutand then reduced to 8 lm thickness by mechanical grindingand polishing. Final thinning was done by ion milling.

Fig. 1. Specific mass = 2 ion current I2 and mass loss Dm vs. temperaturefor as-received and pre-oxidized TiH2 powders heated at 10 K min-1 inflowing Ar.

3. Results

3.1. Phases before heating

As-received TiH2 is cubic, belonging to the F m�3m spacegroup (s.g.) in which the Ti atoms form a face-centredcubic (fcc) arrangement and the hydrogen atoms sit onthe tetrahedral sites. The diffractogram is compatible withTiH1.924 (PDF #25-0982 of the ICDD database) and theRietveld-refined parameter is a = 4.4509 ± 0.0004 A.

After oxidation pre-treatment, the hydride peaks moveto higher 2h values and peaks of TiO2 rutile and Ti3Oappear. The refined lattice parameter of the hydride afterpre-treatment is a = 4.4171 ± 0.001 A. TiO2 rutile belongsto the tetragonal s.g. P42/mnm with lattice parametersa = 4.593 A and c = 2.958 A (PDF #75-1757). Ti3O is tri-gonal (PDF #72-1806), its s.g. P3�1c is and its lattice param-eters are a = 5.15 A and c = 9.56 A, respectively.

3.2. Mass loss and H2 release vs. temperature

Mass loss Dm and specific mass = 2 ion current (i.e.mass = 2 ion current divided by the initial sample mass)I2 vs. temperature for as-received and pre-oxidized TiH2

powders are given in Fig. 1. I2 is a measure of the H2

release rate, whereas the mass loss is the correspondingcumulative value. Both curves provide complementaryinformation, but for simplicity hereafter we refer to the I2

curve as the H2 release curve. Dm is used to determinethe H concentration (CH) in the powder. Onset tempera-tures of H2 release are defined by two intersecting tangentsas shown in the inset of Fig. 1. As-received TiH2 starts torelease H2 from 220 �C followed by a small peak at330 �C, accompanied by negligible mass loss only. Abovethe onset temperature of 375 �C, mass loss regimes are cor-related with peaks of H2 release at 439, 486 and 607 �C.After a kink observed at 780 �C, the H2 release slows downand ceases at 960 �C. The total mass loss from the as-received powder is 3.8 wt.%. Pre-oxidized TiH2 has a firstonset at 461 �C and a second at 535 �C followed by a singlepeak at 647 �C, after which H2 evolution declines to end atabout 890 �C. The total mass loss from the pre-oxidizedpowder is 2 wt.%.

3.3. Phase transformation sequences

3.3.1. As-received powder

Results from the in situ ED-XRD experiments are givenas density maps of intensities containing the diffractionlines of phases dispersed in energy vs. time in Fig. 2. In

Fig. 2. Density maps of diffracted intensities (colour scale) dispersed in energy (in keV) vs. time (in min) show the evolution of the d, b and a diffractionlines in as-received powder while applying the temperature profile TH (t) under: (a) flowing Ar and (b) stationary Ar plus the released H2. Number triplesare (h k l) indices. Legends connected to the measured sample temperature profile TS(t) indicate corresponding figure and sample condition for TEMinvestigations. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)


Fig. 2a, the diffraction lines of the d, a and b phases in as-received powders evolve in time and temperature under Arflow. Legends “first” and “last” indicate at what tempera-ture phases appear and vanish. The starting phase is d.As temperature increases the two hydrogen-containingsolid solutions a and b appear. The a phase is hexagonalclose-packed (hcp, s.g. P63mmc), b is body-centred cubic(bcc, s.g. Im�3m) and hydrogen atoms sit randomly on thetetrahedral sites of both phases [19]. For simplicity, weoften refer to these three phases as d, a and b. The sampletemperature TS(t), measured at the height of the diffractingvolume located about 1.5 mm above the heating plate,deviates from TH(t) because heat conduction through thepowder samples deteriorates with increasing temperature.Moreover, the Ar flow enhances the deviation of TS(t) fromTH(t) since TSmax is 807 �C under Ar flow but 880 �C in astationary Ar atmosphere (compare Fig. 2a with b). Evenif we acquire one spectrum every 2.9 K, systematic temper-ature differences of up to 15 K are detected for TS at agiven TH on other samples. Temperatures indicated in thediffraction lines are extracted from the TS(t) profile. Thusthe phase transformation sequence extracted from Fig. 2acan be expressed as:

heating : d��!432 �Cdþa��!514 �C

dþaþb��!536 �Caþb��!702 �C

a

cooling : no phase transformation ða remainsÞ

(ð1Þ

Sequence (2) is obtained, also under Ar flow but afterreducing the heating rate from 10 to 5 K min�1. The effectis a shift by 27, 58, 44 and 59 K towards lower tempera-tures for each phase transformation.


dþaþb��!492 �Caþb��!643 �C

a

cooling : no phase transformation ða remainsÞ

(

ð2ÞSequence (1) gets strongly modified after changing

the atmospheric conditions from flowing to stationaryAr. In this case the phase transformation extracted fromFig. 2b is:


dþaþb��!620 �Caþb��!744 �C

b

cooling : b��!568 �Cbþa��!264 �C

bþaþd��!157 �Caþd

(

ð3ÞThe first three phase transformations of sequence (3)

are shifted to higher temperatures compared to theirequivalents in sequence (1). However at 744 �C, a trans-formation from (a + b) into b occurs and during coolingthree phase transformations resulted in the mixture(a + d).

The phase transformation sequence followed by AD-XRD (see Supplementary Fig. B) is essentially the sameas sequence (1), differing only in terms of temperature asfollows:

heating : d ��!440 �Cdþ a ��!465 �C

dþ aþ b ��!478 �Caþ b ��!620 �C

an

ð4ÞIn the phase transformation followed by neutron diffrac-

tion on TiD2, d is the starting phase at 348 �C. It trans-forms between 633 and 656 �C into b which remained up

Fig. 3. Same as Fig. 2a but for pre-oxidized TiH2 powder. Peaks of Ti3Oand TiO2 are indicated by o and r, respectively. The symbol “*” refers tothe compositional modification after pre-oxidation, but d*, b* and a* arestructurally isomorphous to d, b and a.


to 897 �C. After cooling down to 277 �C, d is again the onlypresent phase (see Supplementary Fig. C).

3.3.2. Pre-oxidized powder

Fig. 3 is the density map of diffracted intensities andtemperature profiles for the pre-oxidized powder. Theasterisk “*” in d*, a* and b* denotes compositionally mod-ified phases caused by the pre-oxidation, i.e. containingmore O and less H. Apart from this, the crystal structuresare the same as of d, a and b. This sample initially containsthree phases: d*, Ti3O and TiO2. Lines of Ti3O match theones of a* (see e.g. O112 connected to a�101 in Fig. 3). Thetransition from Ti3O into a* at 500 �C is assumed because

Fig. 4. (a) SEM image showing the morphology of as-received powder. (b)sequence of diffraction rings confirms that d-TiH2 phase is the only phase pre

according to the Ti–O binary phase diagram Ti3O is notstable above this temperature [20,21]. Up to 500 �C thelines of all phases move towards lower energies, indicatinglattice expansion, after which at 538 �C b* appears and thelines of all phases shift towards higher energies, indicatinglattice contraction. The diffraction lines of TiO2 fade atabout 623 �C (see r101 line). b* disappears at 673 �C anda* is the final phase. The phase-transformation sequenceof pre-oxidized TiH2 can therefore be expressed as::

heating : d� þ Ti3Oþ TiO2 !500 �C

d� þ a� þ TiO2

!538 �Cd� þ a� þ b� þ TiO2 !

623 �Cd� þ a� þ b�

!630 �Ca� þ b� !673 �C

a�

cooling : no phase transformation ða� remainsÞ

8>>>>><>>>>>:

ð5Þ

3.4. Powder morphology and spatial distribution of phases

3.4.1. As-received powderThe condition for the as-received powders is indicated in

the sample temperature profile TS(t) in Fig. 2a as “beforeheating”. Its morphology is shown in Fig. 4. The powderparticles have flat surfaces, sharp edges and a broad sizedistribution which ranges from micron-sized particles(Fig. 4a) down to particles of about 100 nm diameter(Fig. 4b). The ring-shaped SAED pattern obtained forthe region shown in Fig. 4b indicates that the as-receivedTiH2 particles are polycrystalline. All rings are indexedby dhkl values of the d phase only.

The initial microstructure evolves to the one shown inFig. 5 after cooling from 430 �C (see TS(t) in Fig. 2a).Fig. 5a is a ZLF bright-field image of a larger particle slicedfor TEM. The microstructure of the central particle com-prises two different parts: a homogeneous core and anouter shell consisting of small crystallites, which accordingto the indexed SAED patterns in Fig. 5b has the structureof the a phase. The core of the particle is single-crystalline dphase as confirmed by its [2 0 0] zone axis pattern (Fig. 5c).

ZLFBF image and SAED pattern for the visible area. The fully indexedsent in this sample.

Fig. 5. (a) ZLFBF image of an as-received TiH2 particle after heating to 430 �C and cooling (see also Fig. 2a). (b) Indexed SAED pattern obtained fromthe indicated region giving evidence of an a-phase shell. (c) SAED from the core corresponds to a [2 0 0] zone axis of the d phase.

Fig. 6. (a) SEM image of as-received TiH2 after heating to 807 �C and cooling, showing sintering of the powder particles, porosity and grain coarsening.(b) ZLFBF image in which triple points of grain boundaries are indicated by arrows. The diffraction pattern corresponds to the [1 1 0] zone axis of the aphase.


After heating to 807 �C and subsequent cooling, the sam-ple microstructure changes to the one shown in Fig. 6. TheSEM image (Fig. 6a) shows sintering of the former powderparticles, grain growth and residual porosity. Triple pointsof grain boundaries are marked by arrows in the bright-fieldTEM image (Fig. 6b). The crystallographic structure ofthese large grains is identified as belonging to the a phaseby electron diffraction (inset of Fig. 6b).

3.4.2. Pre-oxidized powder

Fig. 7a shows a cross-sectioned particle in the condition“before heating” as indicated in Fig. 3. Thickness varia-tions lead to varying contrast in the core. A small regionof uniform contrast in the core allowed us to performSAED and identify the d* hydride phase (Fig. 7b). Accord-

ing to the ring-shaped SAED pattern shown in Fig. 7c, theshell is a mixture of Ti3O and TiO2 oxides. The oxygenenrichment in the shell compared to the core is corrobo-rated by the oxygen map given in Fig. 7d. After heatingto 805 �C and cooling, see TS(t) in Fig. 3, this sampleremains a powder (Fig. 8a), i.e. no sintering takes place.The SAED pattern shown in Fig. 8b corresponds to theindicated region and is a [0 0 2] zone axis of the hcp a*

phase.

3.5. Oxygen content analysis

The oxygen content analysis of samples “before heat-ing” and “after cooling from TSmax” is summarized inTable 1. These values are the average of three measure-

Fig. 7. Spatial distribution of phases in pre-oxidized TiH2 “before heating” (i.e. the state indicated in Fig. 3) as investigated by TEM. (a) ZLFBF image ofa particle slice. (b) SAED pattern from the core region showing the [2 1 1] zone axis of the d* hydride. (c) SAED pattern of the shell region, where the maindiffracting phases were identified as Ti3O and TiO2. (d) Grey-scale oxygen map of the same area. Lighter grey in the shell indicates oxygen enrichment.

Fig. 8. (a) SEM image of a pre-oxidized TiH2 particle after heating to 805 �C and cooling (see Fig. 3). (b) ZLFBF image showing the outer region of oneparticle on which SAED was obtained. The SAED pattern shows the zone axis [0 0 2] of the hcp a* phase.


ments. The statistical scatter between measurements isabout 3% but systematic errors can be larger. These aver-age values indicate oxygen uptake after heating to TSmax,but the effect could be within the experimental error.

4. Discussion

4.1. Mass loss and H2 release

The observation of a multi-peak structure of H2 releasefrom the as-received powder and of a single peak for pre-oxidized powder as shown in Fig. 1 confirms previousreports by other authors applying similar experimentalconditions [3,7–10]. The total mass loss of 3.8% from the

as-received powder agrees with reported values that rangefrom 3.5 to 3.8 wt.% [3,8,9,22]. In the present study, theonset of the first prominent peak of H2 release occurs at375 �C, whereas the cited studies report onset temperaturesbetween 380 and 450 �C, though for coarser powders (par-ticle size 50–63 lm) than the one used in this work (particlesize <36 lm). Bhosle et al. [23] and von Zeppelin et al. [24]showed that if the particle size of TiH2 is reduced by ballmilling, which is precisely the method for reducing the par-ticle size of these powders [25], H2 release starts at lowertemperatures.

Hydrogen release from the pre-oxidized TiH2 yields a2 wt.% mass loss, i.e. less than the values of 2.3–2.5 wt.%reported for the same oxidation pre-treatment [8–10]. This

Table 1Measured oxygen content of TiH2 powder samples.

Oxygen content, wt.%

Before heatinga After coolinga

As-received 1.4 1.7Pre-oxidized 8.7 9.3

a Conditions indicated in Figs. 2a and 3.

Table 2Calculated diffusion coefficients of H through b, a and TiO2 at 500 and 600 �

Diffusion coefficientsa 500 �C 600

DbH; cm2 s�1 7.56 � 10�5 9.87

DaH; cm2 s�1 4.05 � 10�6 1.06

DTiO2

H ; cm2 s�1 2.56 � 10�7 7.06

a Calculated using pre-exponential factors and activation energies from the

Tem

pera

ture

°C

H, atTi 0 10 20 30

900

800

700

600

500

400

300

200

100

α

300

6.7

882

1

H, w

a

b

Fig. 9. (a) Phase diagram of Ti–H system adapted from Ref. [19]. 3.5 wt.% H atentering the d + b field is discussed in the text. (b) Radial distribution of phaparticles comprise the fv volume fractions of phases. The H2 release curve (sam


difference can be related to the oxygen content of the pow-ders. For our finer powder the oxygen content increasesfrom 1.4 to 8.7 wt.% during pre-oxidation (Table 1), andthe thickness of the resulting oxide layer is about 100 nm(Fig. 7). Matijasevic et al. reported an increase from 1.0to 6.0 wt.% O for the same pre-oxidation but on coarserpowders and also an oxide shell 100 nm thick [9]. Our finerpowder has a larger surface-to-volume ratio and hence a

C.

�C Ratio Db or aH =DTiO2

H

at 500 �C, at 600 �CRefs.

� 10�5 295, 139 [38]

� 10�5 16, 15 [36]

� 10�7 – [35]

indicated references assuming an Arrhenius law.

ββ

. %

40 50 60 70

39 51.2

δ

2 3 4

H

t. %

T=432°C3.4 wt.% H

432 �C corresponds to the mass loss of 0.4% indicated in Fig. 1. The arrowses vs. temperature applying a core–shell model, and assuming sphericale as in Fig. 1) is superimposed.


higher oxygen content. Because of this, a larger amount ofH2 is sacrificed during pre-oxidation, leading later to lessmass loss during subsequent heating under Ar flow. Theoxide shell acts as an effective diffusion barrier, retardingthe onset temperature due to the lower diffusivity of Hthrough TiO2 (rutile) compared to a and b. Calculated dif-fusion coefficients of H through a and b are 16 and 295times larger than through TiO2 at 500 �C, and 15 and 139times larger at 600 �C, respectively, as given in Table 2.

4.2. Phase-transformation sequence and core–shell model for

as-received TiH2

4.2.1. Apparent “incompatibilities” between sequence (1)and Ti–H phase diagram

Even though the complete dehydrogenation of d led toa, as one expects from the Ti–H phase diagram given inFig. 9a, two features of sequence (1) are incompatible withthe phase diagram. The first one is that at 432 �C, aappears even though the corresponding mass loss is 0.4%(see Fig. 1) or equivalently, 3.4 wt.% H remains in the solid.

10327

527

727

927

1127

1a

b

Fig. 10. (a) Ti-rich part of the O–Ti binary phase diagram adapted from Rediscussed in the text. (b) As Fig. 9b but for a pre-oxidized particle.

This temperature-composition point lies inside the d field.Therefore, according to Fig. 9a there should be only d.As the temperature further increases and the hydrogen con-centration CH decreases due to H2 release, not a but bshould appear. The second apparent incompatibility is thatno coexistence of a + b + d is possible according to thiseutectoid Ti–H phase diagram, but it is neverthelessobserved.

The reproducibility of sequence (1) was tested by heat-ing two more samples up to 611 and 620 �C, respectively.The first three transformations of sequence (1) and theapparent incompatibilities are reproducible within a scatterof 15 K. The temperatures in sequence (1) differ onlyslightly from those of a sequence followed in situ by Mat-ijasevic et al. for coarser TiH2 powder under He flow byAD-XRD at the experimental station KMC-2, and alsoat BESSY II [26], where we obtained sequence (4) for ourfiner powder, which leads to the same result: a appearsbefore b, and d + a + b coexist.

Fig. 9a is the temperature-composition representationfor a highly pure solid in equilibrium with 1 bar of pure

20 30

5 10

f. [21]. Dashed arrows and particular temperature and compositions are


H2 gas [19]. In addition to the influence of the atmosphere,impurities and heating rate can cause significant deviationsfrom the Ti–H phase diagram [19,20,25]. Indeed, by reduc-ing the heating rate from 10 to 5 K min�1, sequence (2), allphase transformations are shifted towards lower tempera-tures, indicating that the heating rate of 10 K min�1 putthe system out of thermodynamic equilibrium.

4.2.2. a before bRemoval of the released H2 gas from the surrounding

atmosphere by Ar flow is one of the reasons for a appear-ing before b. Due to Ar flow the initial H2 partial pressureof the surrounding atmosphere (PH2) is very small, creatinga large hydrogen concentration gradient between solid andgas and favouring the formation of the phase with the low-est H solubility. Within the temperature range studied, theH solubility in a (Sa

H) is the lowest among the three phases,i.e. Sa

H < SbH < Sd

H (see Fig. 9a). Obviously, a is the phasemost compatible with the atmospheric conditions. In addi-tion, the initial oxygen content in the powders (Table 1)and possible oxygen traces in the atmosphere favour a for-mation since O is an a stabilizer [21] (see Fig. 10a). The lar-ger surface-to-volume ratio of a powder sample comparedto bulk amplifies the effect of Ar flow, which in turn pro-motes the persistence of a throughout decomposition.

A modification of the atmosphere proves the effect of Arflow. Sequence (3) is obtained by not letting Ar flow. In thisexperiment, a also appears prior to b because the initialPH2 was zero, but as the decomposition proceeds, PH2 ofthe surrounding atmosphere increases and hence theback-reaction H2(gas) 2H(in the solid) is enabled. This hindersfurther H2 release as reported by Schonefelder and Swisher[12]. In sequence (3), a disappears at 744 �C because PH2 isno longer zero and the remaining H in the solid stabilizes bup to the maximum temperature. b is a high-temperaturephase which in the Ti–H system cannot be retained by nat-ural cooling, hence it transforms into a + d.

One more example is the phase transformation of TiD2

under initially resting Ar atmosphere. The applied temper-ature profile is the same as the one given in Fig. 2b forTiH2, but much larger TiD2 samples than TiH2 were used(3 g vs. 200 mg). Because of a larger sample, the back-reac-tion (i.e. the stabilization of b) is established sooner. Theresult is that the deuteride decomposes less than thehydride. In this more extreme case the sequence for TiD2

is d ��!heatingb ��!cooling

d and a was not detected at all.

4.2.3. d + a + b coexistence

The coexistence of a + b + d even at a reduced heatingrate of 5 K min�1 suggests that decomposition is a hetero-geneous process due either to a CH gradient inside the par-ticles and/or to a variable PH2 depending on the particleposition in the sample. The CH gradient within the parti-cles—lower at the surface in the a shell than in the d/bcore—is linked to the H solubility ranking Sa

H < SbH < Sd

H:A variable PH2 for different particle positions in the

sample might arise from a reduced rate of H2 removal

for particles inside the sample compared to particles atthe sample surface. This effect is demonstrated by sequence(4) probing close to the sample surface where H2 removalwas most efficient. In sequence (4) d + a + b coexist overa temperature range of 13 K compared to 22 K fromsequence (1) referring to the volume. This indicates thatPH2 still varied depending on the particle position withinthe thin surface layer probed. However, this variation islinked to the volume probed and the corresponding hydro-gen concentration profile within.

4.2.4. Core–shell model for as-received powder

In a core–shell model describing the decomposition pro-cess of as-received powder, the a phase appearing duringheating should form a shell around a d core and subse-quently remain at the surface as discussed in Section4.2.2. Upon further heating, b can be assumed to forman intermediate layer between a and d consistent with theCH gradient inside the particles (see Section 4.2.3). A tem-perature-dependent core–shell model is given in Fig. 9b asa radial diagram of phases. In this model, we assume aspherical particle of 3 lm radius (experimental diameterD50 = 6 lm) and ignore possible two-phase regions. Vol-ume fractions of phases fvol,i are calculated as describedin Appendix A. In this core–shell representation, the thick-ness of the a shell initially grows at the expense of the dcore. Later, b appears and grows at the expense of bothd and a. Afterwards, d disappears and a grows again atthe expense of the b core. The superimposed H2 releasecurve shows that as the a shell becomes thicker, H2 releasetends to decrease. The main peak of H2 release starts fromabout 530 �C when the transition from a d core into a bcore takes place, and the a shell reaches a temporary min-imum thickness but does not vanish. In this way, the ideasuggested previously [10,26,27,32] of a controlling H2

release is compatible with this model because a has the low-est H solubility among the three phases (Sa

H < SbH < Sd

H)and also Da

H < DbH (see Table 2). This arrangement of

phases is consistent with the TEM investigations (Figs.4–6) though it ignores the sintering undergone at high tem-perature. Both the fvol,i calculation and the H2 release curverepresent an average over a volume containing a largenumber of particles. In reality, however, particles at thesurface evolve differently than inside the sample and fineparticles should behave differently compared to coarseones. Nevertheless, the radial diagram reasonably corre-lates H2 release with the phase-transformation sequenceand allows the controlling role of a to be visualized.

4.3. Phase-transformation sequence and core–shell model for

pre-oxidized TiH2

4.3.1. Ti3O transformation into a* and disappearance of theTiO2 shell

The match between the diffraction lines of Ti3O and a* isan important feature in the phase transformation of pre-


oxidized powder. It occurs because Ti3O is a superstructureof a*. In Ti3O, Ti atoms form an hcp arrangement with 1/3of the octahedral sites orderly occupied by O atoms [28].However, Ti3O is stable only up to 500 �C [20] as denotedin the Ti–O phase diagram (Fig. 10a), where a dashedarrow indicates that if the temperature increases to abovethe Ti3O field, the peritectoid transitionTi3O ��!�500 �C

aTiþ Ti2O takes place [20,21]. Ti2O is also asuperstructure of a* and its principal diffraction lines matchthose of Ti3O and a* as well [28]. Above 535 �C the randomsolid solution of O in Ti, a*, becomes the stable phase. Forthis reason, we adopt 500 �C as starting temperature for theoverall transition from Ti3O into a* in Fig. 10b.

The O solubility in a* increases above 535 �C andreaches 33 at.% at 600 �C [21]. This influences anotherevent of the phase transformation: the TiO2 lines inFig. 3 become less intense from 535 �C and vanish at623 �C, suggesting that rutile disappears. It is known fromthe literature that if TiO2 reduces its O content, the diffrac-tion lines become less intense [29]. As the total O content inthe solid does not decrease during heating under Ar (seeTable 1), the O concentration profile that before heatinghad O concentrated at the particle surface (Fig. 7d) evolvesto a uniform distribution in a*. It is known that TiO2 canbe reduced by Ti [30]. This process most probably com-mences around 500 �C when the a* diffraction lines gainintensity, i.e. when fvola� increases. The point (9.3 wt.% O,805 �C) taken from Table 1 and Fig. 3, respectively, inFig. 10a lies inside the a* field, i.e. thermodynamics suggest

Fig. 11. Lattice parameters of d, b, a for (a) as-received TiH2 and (b) d*, b*, a*

and particular temperatures (from Figs. 2a and 3) are superimposed.

that only a* as solid product can be expected after completedehydrogenation.

4.3.2. Core–shell model for pre-oxidized powders

A simplified radial compositional diagram vs. tempera-ture is presented in Fig. 10b for an idealized pre-oxidizedTiH2 particle. Before heating, pre-oxidized TiH2 particleshave three phases arranged in a core–shell structure con-sisting of a TiO2 shell, an intermediate layer of Ti3O anda d* core. The evolution in temperature of this radial dia-gram is consistent with the initial core–shell structure, thephase transformation sequence (5) and the complementaryTEM investigations on pre-oxidized powder. The field andspatial distribution of Ti3O and a* are connected due to thetransition from Ti3O into a* at 500 �C. b* is assumed toform an intermediate layer between d* and a* as in theas-received powder. The thickness of the TiO2 layerdecreases from 535 �C, whereas the a* layer grows not onlyat the expense of TiO2 but also of the (d + b) core until a*

becomes the final phase. The superimposed H2 gas releaseunderlines the controlling role of TiO2 on H2 outgassing asits onset coincides with the point at which the TiO2 shellstarts to become thinner, i.e. 535 �C.

4.4. Correlation between contraction of lattice parameter and

peaks of H2 release

Lattice parameters of the phases in as-received and pre-oxidized TiH2, calculated as described in Appendix A, are

and Ti3O for pre-oxidized TiH2 vs. temperature. The curves of H2 release


given in Fig. 11 jointly with the respective H2 release curvesto emphasize their correlation. The general trend in Fig. 11is that during isochronal heating the lattice parametersincrease due to thermal expansion as long as H2 release isnot prominent and decrease after dehydrogenationbecomes significant.

Below the onset temperature of the as-received powder(375 �C), ad expands linearly (Fig. 11a). The correspondinglinear expansion coefficient of d between ambient tempera-ture and 375 �C is 10.65 � 10�6 K�1 which is in goodagreement with the value 10.80 � 10�6 K�1 reported bySetoyama et al. [33]. In the following, ad contracts in tworegimes correlated with two peaks of H2 release. Similarly,the main H2 release peak results in a monotonic contrac-tion of ab.

In contrast, ca expands temporarily between 500 and550 �C. This could be due to an increase in the oxygen con-centration in the a shell. Since in this period the a shellbecomes thinner (see Fig. 9b), and the total oxygen contentof the sample does not decrease (see Table 2), the a shellbecomes richer in O because a remains the preferred hostfor O [31]. Another effect contributing to the expansionof ca is the slight Sa

H increase above 300 �C (see Fig. 9a)[19]. This argument is also valid for aa but the occupancyof either tetrahedral sites by H or octahedral sites by Omodifies preferentially the distance between c planes in a,which in turn affects ca more significantly than aa [19,21].After this period, between 550 and 710 �C, ca contractsduring the strongest dehydrogenation stage. In fact, aa con-tracts up to 740 �C, after which both aa and ca expand,indicating the dominance of thermal expansion over latticecontraction.

For the pre-oxidized sample, the lattice parameters aa�

and ca� of the hcp a* phase that were calculated for thewhole temperature range represent below 500 �C, fractionsof the lattice parameters of Ti3O determined by the expres-sions: aTi3O ¼

ffiffiffi3p� aa� and aTi3O ¼ 2 � ca� [28]. Thus, the dis-

tinction done in Fig. 11b between Ti3O and a* is in order tobe consistent with the assumed transition from Ti3O into a*

at 500 �C discussed in Section 4.3.1. The smaller initial ad�

compared to ad comes from the lower initial hydrogen con-tent [19]. In contrast to ad, the initial expansion of ad� isnonlinear. The expansion of ad� in this pre-oxidized sampleis shifted by 160 K up to 535 �C with respect to the as-received one, in analogy to the shift of the onset of H2

release. The more modest contraction of ad� and ab� above535 �C compared to their isomorphic ad and ab is due to thelower total H2 released from the pre-oxidized powder.Same as in the as-received powder aa� contracts furtherthan ca� in the pre-oxidized sample.

The difference in absolute values between the latticeparameters of a and a* above 650 �C suggests that thesmaller ca and aa more efficiently restricted H2 outgassingthan ca� and aa� . This could explain the faster decay andexhaustion of hydrogen above 650 �C from the pre-oxi-dized powder controlled by a relatively thick a* shell com-pared to the thinner a shell covering the as-received powder

(compare Figs. 9b and 10b above 650 �C). Additionally,the higher the O content, the more the hcp structure is sta-bilized [21]. Due to the initial oxide layer, the pre-oxidizedpowder remains a loose powder after cooling from 805 �C(see Fig. 8a), which does not affect the effectiveness of H2

removal, whereas the as-received powder undergoes someinterparticle sintering and compaction (see Fig. 6a), whichdelayed the final stages of dehydrogenation.

5. Summary

We correlated the phase-transformation sequences ofboth as-received and pre-oxidized TiH2 powders with theobserved peaks of H2 release. The sequence of dehydroge-nation for pre-oxidized TiH2 is reported in this study forthe first time. For heating at 10 K min�1 under flowingAr only thermal expansion of lattice parameters occursand there is hardly any H2 release up to 375 �C in the as-received powder and up to 535 �C in the pre-oxidizedone. Upon further temperature increase, with every peakof H2 release, the lattice parameters contract.

Based on complementary TEM analysis we validatedexperimentally coreshell models for describing the hetero-geneous decomposition processes of both as-received andpre-oxidized TiH2 powders. These models imply the pres-ence of an a shell controlling H2 outgassing from the as-received TiH2 powder, whereas a TiO2 rutile shell controlsthe outgassing from the pre-oxidized TiH2 powder.

The flow conditions during heating were shown to beessential by carrying out various experiments at differentconditions in which the evolving hydrogen was eitherremoved continuously, stabilizing a, or allowed to form asurrounding hydrogen atmosphere that stabilized b at themaximum temperature.

Acknowledgements

We thank M. Mukherjee, Ch. Abromeit, N. Wanderka,S. Klaumunzer and Ch. Genzel for constructive discus-sions, Ch. Forster for helping with TEM sample prepara-tion and H. Kropf with SEM. O. Goerke from TU Berlinassisted the oxygen content analysis.

Appendix A. ED-XRD data analysis

Bragg’s law reads in its energy-dispersive form as:

Ehkl ¼ hc=2dhkl sin h; ðA:1Þwhere h is Planck’s constant and c the speed of light [15,16].The diffraction lines d220, b200, a101, a102 and r101 (of TiO2)fitted by Gaussians allow dhkl values to be calculated viaEq. (A.1). These lines also provide estimates for the volumefractions of phases fvol,i, with i = d, d*, a, a*, b, b*, Ti3Oand TiO2, by dividing the integrated intensity of the princi-pal line Ii by the summed integrated intensity Isum of theprincipal lines of all the phases present at a given tempera-ture [37]. Each volume fraction is given by the relation, fvo-


l,i = Ii/Isum. Ia and Ia� are directly obtained from the princi-pal lines a101 and a�101. The latter is considered o112 (belong-ing to Ti3O) below 500 �C, thus rendering ITi3O.

For the d and d* phases, we integrate the lines d220 andd�220 whose fractional relative intensity is 0.57 with respectto the principal line, i.e. Id or d� ¼ Id220 or d�220

=0:57:Likewise, Ib or b� ¼ Ib200 or b�200

=0:15 and ITiO2¼ I r 101

=0:57.Lattice parameters are calculated using the following rela-tions adapted from Ref. [34]:

ad or d� ¼ 8 � d220;d or d� ; ab or b� ¼ 4 � d200;b or b� ;

aa or a� ¼ d101;a or aa� � d102;a or a�=ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffid2

102;a or a� � d2101;a or a�=4

q; and

ca or a� ¼ffiffiffi3p� d101;a or a� � d102;a or a�=

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffid2

101;a or a� � d2102;a or a�

q:

Appendix B. Supplementary material

Supplementary data associated with this article can befound, in the online version, at doi:10.1016/j.actamat.2011.06.042.

References

[1] Banhart J. Prog Mater Sci 2001;46:559.[2] Erk KA, Dunand DC, Shull KR. Acta Mater 2008;56:5147.[3] Liu H, He P, Feng JC, Cao J. Int J Hydrogen Energy 2010;34:3018.[4] Duarte I, Banhart J. Acta Mater 2000;48:2349.[5] Matijasevic B, Banhart J. Scripta Mater 2006;54:503.[6] Gergely V, Clyne TW. Adv Eng Mater 2000;2:175.[7] Kennedy AR. Scripta Mater 2002;47:763.[8] Lehmhus D, Rausch G. Adv Eng Mater 2004;6:313.[9] Matijasevic-Lux B, Banhart J, Fiechter S, Gorke O, Wanderka N.

Acta Mater 2006;54:1887.[10] Malachevsky MT, D’Ovidio C. Scripta Mater 2009;61:190.[11] Sandim HRZ, Morante BV, Suzuki PA. Mater Res 2005;8:293.[12] Schoenfelder CW, Swisher JH. J Vac Sci Technol 1973;10:862.[13] Gruner WJ. Analytical Chem 1999;365:597.[14] <http://www.ill.eu/sites/fullprof/>.

[15] Genzel Ch, Denks I, Gibmeier J, Klaus M, Wagener G. Nucl InstrumMeth Phys Res A 2007;578:23.

[16] Denks I, Ch Genzel. Nucl Instrum Meth Phys Res B 2007;262:87.[17] <http://www.certif.com/spec.html>.[18] <http://www.helmholtz-berlin.de/userservice/neutrons/instrumenta-

tion/neutron-instruments/e6_en.html>.[19] San Martin A, Manchester FD. Bull Alloy Phase Diag 1987;8:30.[20] Kubaschewski-von Goldbeck O. Phase diagrams. In: Komarek,

editor. Titanium: physico-chemical properties of its compounds andalloys. Vienna: IAEA; 1983.

[21] Landolt-Bornstein New Series IV/5 MSIT�, O–Ti (Oxygen–Tita-nium). New York: Springer Materials; 2005.

[22] Proa-Flores PM, Drew RLA. Adv Eng Mater 2008;10:830.[23] Bhosle V, Baburaj EG, Miranova M, Salama K. Mater Sci Eng A

2003;356:190.[24] von Zeppelin F, Hirscher M, Stanzick H, Banhart J. Compos Sci

Technol 2003;63:2293.[25] Muller WM. Titanium hydrides. In: Muller WM, Blackledge JP,

Libowitz GG, editors. Metal hydrides. New York: Academic Press;1968.

[26] Matijasevic B. PhD thesis, Technische Universitat Berlin; 2006.[27] Takasaki A, Furuya Y, Ojima K, Taneda Y. J Alloys Compd

1995;224:269.[28] Kornilov II, Vavilova VV, Fykin LE, Ozerov RP, Solowiev SP,

Smirnov VP. Met Trans 1970;2:2569.[29] Meyer RJ, Pietsch EHE. Gmelins Handbuch der Anorganischen

Chemie: Titan, vol. 41. Weinheim: Verlag Chemie; 1951.[30] Wiberg E, Wiberg N, Holleman AF. Holleman-Wiberg Inorganic

Chemistry, 34th ed. New York: Academic Press; 2001.[31] Schlapbach L. Surface properties and activation. In: Schlapbach L,

editor. Hydrogen in intermetallic compounds II. Berlin: Springer-Verlag; 1992.

[32] Swisher JH. Sandia Laboratories Report No. SCL-RR-720059; 1972.[33] Setoyama D, Matsunaga J, Muta H, Uno M, Yamanaka S. J Alloys

Compd 2004;381:215.[34] Cullity BD. Elements of X-ray Diffraction. Reading, MA: Addison-

Wesley; 1978.[35] Hill GJ. Brit J Appl Phys (J Phys D) 1968;1:1151.[36] Kubaschewski O. Diffusion. In: Komarek, editor. Titanium: physico-

chemical properties of its compounds and alloys. Vienna: IAEA;1983.

[37] Laine ESU. J Phys F: Metal Phys 1978;8:1343.[38] Matzke HJ, Rondinella VV. Diffusion in carbides, nitrides, hydrides

and borides. In: Beke DL, editor. Landolt-Bornstein group III, vol.33B1. New York: Springer Materials; 2010.



http://www.ill.eu/sites/fullprof/

http://www.certif.com/spec.html

http://www.helmholtz-berlin.de/userservice/neutrons/instrumentation/neutron-instruments/e6_en.html

http://www.helmholtz-berlin.de/userservice/neutrons/instrumentation/neutron-instruments/e6_en.html

decomposition of tih2 studied in situ by synchrotron x-ray and neutron diffraction

Documents