Structural properties of the nanoscopic Al85Ti15 solid solution observed in the hydrogen-cycled NaAlH4 + 0.1TiCl3 system

Available online at www.sciencedirect.com

Acta Materialia 56 (2008) 4691–4701 www.elsevier.com/locate/actamat

Structural properties of the nanoscopic Al85Ti15 solid solution observed in the hydrogen-cycled NaAlH4 + 0.1TiCl3 system M.P. Pitt a,*, P.E. Vullum b, M.H. Sørby a, M.P. Sulic c, C.M. Jensen c, J.C. Walmsley b, R. Holmestad b, B.C. Hauback a b

a Institute for Energy Technology, PO Box 40, Kjeller N-2027, Norway Department of Physics, Norwegian University of Science and Technology, N-7491 Trondheim, Norway c Department of Chemistry, University of Hawaii, Honolulu, HI 96822, USA

Received 11 March 2008; received in revised form 14 May 2008; accepted 17 May 2008 Available online 27 June 2008

Abstract The twice-hydrogen-cycled NaAlH4 + xTiCl3 (x < 0.15) system has been studied by high-resolution X-ray synchrotron diffraction and transmission electron microscopy. Intense low d-spacing shoulders are formed on Al reflections, indicating the formation of a face-cen˚ . The Al85Ti15 solid solution is found as isolated 4–25 nm nanocrystaltred cubic Al85Ti15 solid solution of unit cell dimension 4.0356 A lites on the NaAlH4 surface. The Al85Ti15 phase is highly h1 1 0i{1 1 1} edge dislocated to 6  1016 m2. Local energy-dispersive spectroscopy shows an Al:Ti ratio consistent with an Al85Ti15 composition, confirming an extended solubility of Ti in Al. Structural analysis indicates a quenched L12 superlattice obtained by Al and Ti sublattice swapping, yielding an equivalent A1 Fm 3m model to describe the Al85Ti15 crystal structure. Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Nanocrystalline materials; Titanium aluminides; Solid solution; Crystal structure; Dislocation structure

1. Introduction The binary Al–Ti system has been of great interest to the metallurgical community as the source of potential lightweight materials for aerospace applications [1]. Much work has focused on the Al3Ti phase, which possesses attractive properties such as a low, 3.3 g cm3 density, a high hardness with a yield stress of 980 MPa [2], and a high melting temperature, 1400 °C. It has also proved an interesting system in which to study the temperature-dependent superstructure symmetry interconversion of the D022, D023 and L12 structure types [3]. Due to a lack of ductility below 620 °C [4], studies have focused on converting the tetragonal D022 phase to the cubic L12 phase by partial substitution with transition metals [2,5,6], giving a ternary cubic L12 phase with a high number of accessible slip systems. *

Corresponding author. Tel.: +47 6380 6087; fax: +47 6381 0920. E-mail addresses: [email protected], [email protected] (M.P. Pitt).

Early models of B substitution in Ni3Al [7,8] indicated enhanced ductility was due to B substituting on the unoccupied body centre of the unit cell and subsequently reducing the covalent directionality of the face centred Al atoms. However, channelling-enhanced microanalysis [9,10] of 0.5 at.% (Zr, Hf, W) L12 Cr-stabilized Al3Ti indicated clearly that the body centre remains unoccupied. Other studies [11] have focused on developing duplex-type microstructures, where dislocation sources are activated at the interface between precipitates in the L12 Al3Ti matrix, such as Al2Ti or AlTi precipitates, or by dispersing Al3Ti precipitates within a more ductile Al matrix [12,13]. For Al(1x)Tix with x < 0.25, there exists no known stable phase in the binary phase diagram [14]. Early work on quenched melt spins stated the solubility limit for Ti in Al as 2 at.% [15]. In later studies, this solubility limit has been quoted as 4 at.% [16], 10 at.% [17], and even up to 27 at.% [18] in sputter-deposited films. By comparison, other Al based Al(1x)Mx (M = metal) solid solutions have clear sol-

1359-6454/$34.00 Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2008.05.037

4692

M.P. Pitt et al. / Acta Materialia 56 (2008) 4691–4701

ubility limits [19]. Some solubility limits for Ti in Al are based on the observation of apparently single-phase laboratory source X-ray diffraction patterns [16,17]. However, such typically medium-resolution data can disguise the presence of multiple phases, and Rietveld modelling of the Al:Ti ratio is inherently correlated with the scale factor. Although a partially ordered L12 structure is suggested for x < 0.25 by several authors [13,16], clear quantitative information on the local Al:Ti ratio obtained by a supporting spectroscopic technique such as electron-dispersive spectroscopy (EDS) is typically not provided. Further, there exist no diffraction patterns in the literature showing clear L12 superlattice reflections, and all presented diffraction patterns appear with Fm 3m type symmetry. The correct structure and maximum solubility of Ti for Al(1x)Tix with x < 0.25 remains an open question. Two different diffraction studies provide key pieces of information about the likely Al(1x)Tix crystal structure. The first study [13] shows clear L12 superstructure reflections for a single-phase Al90Zr10 sample. This study also states that Al90Ti10 and Al90Hf10 possess similar L12 structure, although no diffraction patterns are shown. The second study [3] provides the variation of unit cell dimension of Al(1x)Tix as a function of Ti content for an assumed cubic unit cell. The strong variation of unit cell dimension with Ti in [3] indicates that Ti continues to be solved into Al up to 25 at.%. Based on Zr and Ti residing in the same group of the periodic table, we then expect similar bonding behaviour and a similar structure for Ti; however, the absence of superlattice reflections requires an explanation. The NaAlH4 system is primarily of interest as a hydrogen storage material [20]; however, the formation of isolated nanoscopic crystalline Al(1x)Tix (x < 0.25) phases on the NaAlH4 surface provides an interesting new test bed for the Al–Ti system at the 4–25 nm scale. In this work we combine high-resolution X-ray synchrotron diffraction and transmission electron microscopy (TEM) to show that the Al85Ti15 phase displays a lattice parameter and Al:Ti ratio in excellent agreement with the predicted 85:15 phase in Ref. [3]. The absence of L12 superlattice reflections for Al85Ti15 is explained with structure factor calculations for L12 sublattice exchange of Ti and Al which quench the superlattice reflections completely. Similar sublattice exchange in the Al(1x)Zrx system reproduces the observed diffraction data in Ref. [13] which shows the Zr system retaining the L12 superlattice reflections, as expected due to Zr possessing a larger X-ray scattering factor than Ti, and with much less Zr solved into the Al sublattice. This paper is organized as follows. Section 3.1 describes the methods used to produce isolated nanocrystals of Al(1x)Tix (x < 0.25) phases, and presents high-resolution images of their dislocation structure and local EDS analysis of elemental Al:Ti ratios. Section 3.2 discusses the potential crystallographic symmetry of such phases, based on model structure factor calculations of sublattice exchange in the Al(1x)Tix (x < 0.25) and Al(1x)Zrx (x < 0.25) structures.

2. Materials and methods NaAlH4 was purchased from Albermarle Corporation (Lot No. 22470404-01). TiCl3 was purchased from Sigma–Aldrich Chemicals Inc. (>99.99% purity). At all times, all powders have been handled under inert Ar atmosphere in a dry glovebox, with <1 ppm O2 and H2O. Milled NaAlH4 powders with TiCl3 additive were prepared in 1 g quantities in a Fritsch P7 planetary mill (PM), with a ball to powder ratio (bpr) of 20:1, at 750 rpm for a period of 1 h, and in 2 g quantities in a Spex 6750 Freezer mill, milled at intensity 15 for a period of 2 h, in a standard SPEX cryo vial, with a 32 g SS440c impactor. Hydrogen (H) cycling was performed in a Sieverts apparatus, rated to 200 bar and 600 °C. Samples for diffraction measurements were removed directly after milling, and after 2 and 5 H cycles (typically H cycled at 140 °C with 150 bar aliquot). Synchrotron radiation powder X-ray diffraction data were recorded at the Swiss–Norwegian Beamline (SNBL) at the European Synchrotron Radiation Facility (ESRF) in Grenoble, France. Samples were contained in rotating 0.8 mm boron–silica glass capillaries. High resolution data (Dd/d 4  104) was typically collected at 295 K between ˚ 5 and 35° 2h, in steps of 0.0025°. A wavelength of 0.4998 A was obtained from a channel-cut Si (1 1 1) monochromator. Medium resolution (Dd/d 3  103) in situ annealing data were collected on a two-dimensional image plate (MAR345) over the 2h range 3–34° with step size 0.015° ˚ was and exposure time of 30 s. A wavelength of 0.7111 A used. TEM was performed with a JEOL 2010F field emission gun operating at 200 kV, or with a Philips CM30 operating at 100–300 kV. TEM samples were loaded inside the glovebox and transferred into the column of the microscope by two different methods. (i) An oxygen-tight transfer cap was used, with the cap being removed inside a glovebag attached to the holder entrance of the microscope. The glovebag was pumped and flushed with pure N2 to prevent sample oxidation. (ii) A Gatan environmental cell TEM holder was used. A vacuum gate valve on the environmental chamber allowed the sample to be withdrawn and isolated in the chamber during transfer, which prevented contamination or contact with air. Method (i) was most frequently used. X-ray synchrotron diffraction patterns were analyzed by the Rietveld method, using the software RIETICA [21]. Diffraction line profiles were fitted with a full Voigt function, with the instrumental shape determined by a NIST LaB6 660a lineshape standard, further annealed to 1800 °C. Backgrounds were modelled with type I Chebyshev polynomials. 3. Results 3.1. Methods to produce nanoscopic Al(1x)Tix (x<0.25) phases from NaAlH4 + xTiCl3 Crystalline Al(1x)Tix x < 0.25 solid solutions can be crystallized from the NaAlH4 + xTiCl3 system by three

(114) D023 (103) D022 (111) L12

(004) D022

6000

(008) D023

7000

(200) D022 (200) D023 (200) L12 NaCl (220) (200) Al

M.P. Pitt et al. / Acta Materialia 56 (2008) 4691–4701

4693

(112) D022

o

502 C o 477 C o 452 C o 426 C o 401 C o 376 C o 350 C o 325 C o 304 C o 275 C o 253 C o 228 C o 203 C o 177 C o 152 C

4000 3000 2000 1000 0 1.90

(220) Na3AlH6 2.00

2.10

(111) Al

Intensity (counts)

5000

(111) Al(1-x)Tix 2.25 d-spacing (Å)

2.30

2.35

2.40

˚. Fig. 1. In situ annealing data to 500 °C from PM NaAlH4 + 0.1TiCl3, across the most interesting range of d-spacing from 1.9 to 2.4 A

4000 3000

o

203 C

1000

(111) Al

2000

(111)

3.1.1. Vacuum annealing In Fig. 1, the changes with temperature of PM NaAlH4 + 0.1TiCl3 are shown by diffraction up to 500 °C. Several phase transitions are clearly evident, with a mixture of L12 and D023 Al3Ti phases present at intermediate temperatures, and by 500 °C, all Ti added to the sample is bound as a mixture of ordered D022 and D023 Al3Ti phases. The comprehensive high-temperature study of Al(1x)Tix phases (Al95Ti5–Al75Ti25) in Ref. [3] demonstrates that if a D022/ D023 mixture has formed at high temperature, it has crystallized from a solid solution in the range Al90Ti10 to Al80Ti20. Formation of the Al(1x)Tix solid solution in our data is evident at 175 °C, indicated as a strong, low d-spacing asymmetry (due to the inherently lower resolution of MAR data) on Al 1 1 1 in Fig. 2. The initially formed Al(1x)Tix solid solution is stable in the temperature range 175– 425 °C. Fig. 3 shows the thermal expansion of Al and

Al(1x)Tix solid solution, and contraction of the L12 unit cell dimensions up to 425 °C. Extrapolating backwards in temperature, it is clear that the Al(1x)Tix solid solution formed by annealing PM-only samples yields an ambient unit cell ˚ , indicating a Al86Ti14 composition, dimension of 4.0375 A based on the Ti-dependent variation in unit cell dimension across the Al to Al75Ti25 range for milled samples in Ref. [3], reproduced with permission as Fig. 4. The high-resolution diffraction data in Fig. 5 shows the development of a broad and intense shoulder on Al 1 1 1 after annealing PM NaAlH4 + 0.1TiCl3 isothermally at 140 °C for 12 h. Indexing ˚ , indicating an yields a cubic unit cell of dimension 4.0397 A

Intensity (counts)

methods: (i) dynamic (heat to >180 °C at 2 °C min1) or isothermal (12 h at 140 °C) vacuum annealing of planetary/ cryomilled (CM) NaAlH4 + xTiCl3 powder; (ii) long-term (>24 h) ambient planetary milling of NaAlH4 + xTiCl3 powder; and (iii) H cycling of planetary/cryomilled NaAlH4 + xTiCl3 powder at 120–140 °C (>2 cycles). In all three cases, after the completion of the milling process, 2–20 nm Al nanocrystals are embedded in an amorphous, Ti-rich Al(1x)Tix (0.3 < x < 0.5) matrix. The nano Al/ amorphous Al(1x)Tix matrix composite is embedded on the NaAlH4 surface [22]. Crystallization of the Al(1x)Tix x < 0.25 solid solution then occurs by local diffusion of Ti from the amorphous matrix into the Al nanocrystals, resulting in isolated 4–25 nm Al(1x)Tix solid solution.

o

0 2.30

Al(1-x)Tix

2.32

2.34 2.36 d-spacing (Å)

177 C

2.38

Fig. 2. Low d-spacing asymmetry on Al 1 1 1 indicating the formation of Al(1x)Tix solid solution by 175 °C. Data are represented by circles (o) and the solid line represents the calculated structure model. The 177 °C pattern does not include Al(1x)Tix solid solution, and significant misfit is observable. The 203 °C pattern includes Al(1x)Tix solid solution, and the asymmetrical misfit is removed.

4694

M.P. Pitt et al. / Acta Materialia 56 (2008) 4691–4701

4.10

4.06 α-Al

Al

4.08

Unit cell cube side (Å) (15hr mill)

4.04

Al(1-x)Tix 4.04

4.02 L12

4.02

4.00

3.98

4.00 0

100

200

300

400

Al3Ti L12

500

Temperature (oC)

3.96 0

Fig. 3. Thermal expansion of Al and Al(1x)Tix solid solution during dynamic vacuum annealing. Backward extrapolation indicates an ambient ˚ for the Al(1x)Tix solid solution, unit cell dimension of 4.0375 A indicating an Al86Ti14 composition.

3.1.3. Hydrogen cycling Fig. 8 shows high-resolution X-ray synchrotron diffraction patterns of twice-H-cycled NaAlH4 + xTiCl3 for x = 0.05, 0.10 and 0.15, in the d-spacing range 2.25–

10

15 Ti at.%

20

25

30

Fig. 4. Variation of Al(1x)Tix unit cell dimension as a function of Ti concentration across the 0–25 at.% Ti range. Reproduced with permission from Journal of Metastable and Nanocrystalline Materials [3].

Al88Ti12 composition based on Fig. 4. Peak breadth analysis shows the crystallites are 17 nm. It is likely that longer isothermal annealing times will yield Ti-richer compositions.

20000

15000

Intensity (counts)

3.1.2. Long-term planetary milling The evolution of the Al(1x)Tix solid solution with longterm planetary milling of NaAlH4 + 0.1TiCl3 is shown by high-resolution synchrotron diffraction in Fig. 6. Intense peak broadening of the Al(1x)Tix solid solution is evident after 48 h. The variation of Al and Al(1x)Tix unit cell dimensions from PM NaAlH4 + 0.1TiCl3 as a function of milling time is shown in Fig. 7. Over 48 h, the Al shows a compression of the unit cell dimension from 4.0490 to ˚ . Based on Fig. 4, this corresponds to Al91Ti9. 4.0432 A ˚, The Al(1x)Tix unit cell dimension decreases to 4.0041 A indicating an Al80Ti20 composition. Compression of the Al unit cell dimension occurs only after long milling times. The 48 h Al80Ti20 composition has an average mosaic size of 17 nm (based on the Scherrer equation), and an approximately doubled strain parameter compared to the Al(1x)Tix solid solution observed in H-cycled powders is easily observed by viewing the considerably less broad shape for typical H-cycled Al85Ti15 in the upper part of Fig. 6. It is clear that by milling for longer than 48 h, the L12 Al75Ti25 unit cell dimension may be reached. Nanoscopic Al75Ti25 (7 nm) can also be obtained in much shorter time by milling the stoichiometric reaction 3NaAlH4 + TiCl3 which reduces completely to 3NaCl + Al3Ti [23].

5

10000

5000 Al (111)

Unit cell (Å)

4.06 Ta

0 Al(1-x)Tix (111) 2.28

2.30

2.32

2.34

2.36

d-spacing (Å) Fig. 5. Isothermally vacuum annealed PM NaAlH4 + 0.1TiCl3. After 12 h at 140 °C, an Al88Ti12 composition has formed with unit cell dimension ˚. 4.0397 A

˚ , immediately around Al 1 1 1. Intense low d-spacing 2.40 A shoulders have clearly formed on the Al reflections. Index˚ , indicating ing gives a cubic unit cell of dimension 4.0356 A an Al85Ti15 composition according to Fig. 4. The linear interpolation across the concentration range 0 < x < 0.25 from ball milling experiments proposed in Ref. [24] is not necessary if Fig. 4 is used, and it is clear that the quantita-

M.P. Pitt et al. / Acta Materialia 56 (2008) 4691–4701

12000

4695

8000 7000

10000

Al (1 -x) Ti x Al (111)

48 hrs

24 hrs

2000

0

(111) Al85Ti15

2.24

2.28

12 hrs

4000

15% Ti Cl 3

3000

2.32 2.36 d-spacing (Å)

2.40

4.05

α-Al to Al91Ti9

4.04

4.03

4.02

4.01

Al87.5Ti12.5 to Al80Ti20 4.00

0

10

20

2000

10% Ti Cl 3

1000

5% TiC l 3

0 2.25

Fig. 6. Formation of the Al(1x)Tix solid solution by planetary milling up to 48 h. Asymmetry on the low d-spacing tail of Al 1 1 1 can be observed as early as 12 h. The top figure compares the lineshape from Al(1x)Tix solid solution in H-cycled powder.

Unit cell cube side (Å)

5000

4000 (111) Al

Intensity (counts)

6000

Intensity (counts)

6000

x5 H cyc.

8000

30

40

50

Milling time (hours) Fig. 7. The variation of Al and Al(1x)Tix unit cell dimensions from PM NaAlH4 + 0.1TiCl3 as a function of milling time up to 48 h. After 48 h the Al(1x)Tix unit cell dimension indicates Al80Ti20.

tive analysis of the PM and H-cycled NaAlH4 + 0.1TiCl3 sample in Ref. [25] yielded an equivalent Al:Ti ratio to this study. Although the Al:Ti ratios are based on the concentration-dependent unit cell variation of Fig. 4, it is not clear if such a ratio can exist in a single Al(1x)Tix type phase, and the effect of the distribution of Al and Ti atoms on

2.28

2.31 2.34 d-spacing (Å)

2.37

2.40

Fig. 8. Formation of the 1 1 1 reflection of Al85Ti15 for 5, 10 and 15% TiCl3 in twice-H-cycled NaAlH4 + xTiCl3. I111 can be observed to increase as a function of the amount of TiCl3 added.

the structure factor of such a phase is investigated in Section 3.2 to ascertain the correct symmetry. NaAlH4 decomposes under a moderate electron flux [26,27]. However, all of the surface-bound Al(1x)Tix phases formed are completely stable under the electron beam, and can be investigated in detail by high-resolution techniques [27]. In addition, EDS or electron energy loss spectroscopy (EELS) can be used to determine the composition and discriminate between Al(1x)Tix and pure Al after the reduction reaction during ball milling or due to decomposition of NaAlH4. Fig. 9 shows a dark-field image of a particle aggregate containing several NaAlH4 grains, from a twice-H-cycled PM NaAlH4 + 0.1TiCl3 sample. The image contrast comes from the Al, Al85Ti15 and Al75Ti25 1 1 1 diffraction signal, where statistically three out of every four particles (measured by quantitative phase analysis (QPA) of the diffraction data) are Al(1x)Tix crystallites. Based on 20 different dark-field images, the size of the Al(1x)Tix particles are found to be in the range 4– 25 nm. Some of the Al(1x)Tix particles are directly exposed at the outer powder surface. However, the direct location with respect to the surface is not conclusive for most of the Al(1x)Tix particles due to the image being a twodimensional projection along the direction of the electron beam. 3.1.4. Dislocation content in the Al(1x)Tix x<0.25 phase A high-resolution image of a typical surface Al(1x)Tix crystallite is shown in Fig. 10a. Lattice fringes, corresponding to (1 1 1) planes, are visible in the image and they extend out to the surface. The high-resolution image in Fig. 10b is a filtered inverse-Fourier-transformed image constructed purely by the elastic Bragg scattered intensity in the Fourier-transformed image. The inelastic scattering has been

4696

M.P. Pitt et al. / Acta Materialia 56 (2008) 4691–4701

Fig. 9. Selected area dark-field image apertured (<150 nm) around Al 1 1 1 of twice-H-cycled PM NaAlH4 + 0.1 TiCl3. QPA of the diffraction data indicates three out of four particles are Al(1x)Tix solid solution, with the dark-field measurements showing them in the size range 4–25 nm.

filtered out to better visualize the lattice and defects. Only line defects along one of the four individual h1 1 1i directions are visible in each projection. In this study, all highresolution images were recorded close to Schertzer defocus conditions. For a sample thickness in the range 4–25 nm, contrast from lattice defects is predominantly due to bulk dislocations extending through most of the sample thickness, in this case h1 1 0i{1 1 1} edge dislocations. Possible surface defects will contribute only very minor contrast in the high-resolution image. Furthermore, contrast from possible surface defects is likely to be removed during the

Fig. 10a. High-resolution image of a surface bound Al(1x)Tix crystallite in planetary milled and five times H-cycled NaAlH4 + 0.1TiCl3.

Fig. 10b. Inverse Fourier transform image from the area marked by the box in Fig. 10a. Only the Bragg scattered intensity has been kept in the Fourier transform before transforming back to real space. Numerous h1 1 0i{1 1 1} edge dislocation cores are evident; several are circled for clarity.

filtering process. Fig. 10c shows a similar image to Fig. 10b, but from NaAlH4 + 0.1TiCl3 which was initially cryomilled and then H-cycled twice. Numerous h1 1 0i{1 1 1} edge dislocation cores are visible (circled). The Al(1x)Tix solid solution is h1 1 0i{1 1 1} edge dislocated to a density of 6  1016 and 7  1016 m2 in PM (Al85Ti15) and CM (Al84Ti16) samples, respectively. The h1 1 0i{1 1 1} dislocation density estimated from the high-resolution Fourier-transformed images of Al(1x)Tix x < 0.25 crystallites can also be modelled within the highresolution synchrotron peakshape. Dislocation lineshape modelling was performed using the full Voigtian implementation [28,29] of the Krivoglaz [30] and Wilkens [31] dislocation lineshape formalism. In every dislocation lineshape analysis of the Al(1x)Tix solid solution in this study, the Gaussian parameter of the model reduced to zero, or very near zero, rendering the quantification of the dislocation density intractable from the synchrotron peakshape. The M parameter indicative of dislocation separation is still quantifiable, and is always low, 0.25, indicating very short separations and dipole-like character. Such short core separations are easily observable on the sub 5 nm scale in the high-resolution TEM images of the Al(1x)Tix solid solution. For the long-term PM NaAlH4 + 0.1TiCl3 sample, commensurate with the strong decrease in Al(1x)Tix unit cell dimension from 24 to 48 h (see Fig. 5) there is a doubling of the Lorentzian dislocation lineshape parameter, indicating that at Ti concentrations higher than 16 at.%, the Al(1x)Tix phase is stabilized by a considerably high density of h1 1 0i{1 1 1}edge dislocations. The dislocation density in the 48 h PM sample is the highest density observed in any

M.P. Pitt et al. / Acta Materialia 56 (2008) 4691–4701

4697

holey carbon film onto which the sample is dispersed. The surface crystallite shown in Fig. 10a was found to have a composition corresponding to Al0.83Ti0.17. Regarding the accuracy of the composition measurements, it should be noted that the k-factor in the Cliff–Lorimer equation [32] was not determined experimentally, and the various thicknesses of the Al(1x)Tix crystallites were only estimated. The value of x given by the EDS software should therefore be treated as semi-quantitative rather than quantitative. EDS spectra were acquired from several different Al(1x)Tix surface crystals, with the value of x varying from 0.13 to 0.18. It is uncertain if these variations are real or partly due to different analyzed thicknesses. In addition, possible amorphous material, having a different composition than the crystalline Al(1x)Tix, might contribute to some of the spectra. However, an approximate value of 0.15 seems to be quite reliable, and we conclude that the Ti solubility in Al extends to at least 15 at.% in crystallites on the 4–25 nm scale. Fig. 10c. An inverse Fourier transform-filtered image of Al84Ti16 in twiceH-cycled, cryomilled NaAlH4 + 0.1TiCl3. h1 1 0i{1 1 1} edge dislocation cores are evident circled.

of our samples, and is estimated at 1  1017 m–2 (based on the doubling of the Lorentzian lineshape parameter, and the estimation of local dislocation density from our high-resolution TEM images for Al85Ti15 in H-cycled powders). 3.1.5. Local measurement of the Al:Ti ratio by EDS In Fig. 11, the EDS spectrum (solid line) from one of the surface crystallites is shown to verify that it is an Al(1x)Tix crystallite, and not pure Al. The dotted line spectrum was acquired from a larger NaAlH4 aggregate and represents more the average composition of the sample. Cu and C are not present in the sample. Cu peaks are present in the spectra due to the Cu grid and the environment around the sample and detector, and the C peak comes from the 4000

Al K

Large area fcc Al1-xTix

Intensity

3000

Na K

2000

1000

Ti Kα CK OK

0

0

1

Cu Kα

Cl Kα Cl Kβ

2

3

Ti Kβ

4

5 keV

3.2. Crystallography and Ti solubility limit of Al(1x)Tix (x<0.25) phases Although the diffraction data from Al(1x)Tix x < 0.25 phases show face-centred reflections only, it becomes difficult to confirm in our data, or in literature data for ballmilled Al(1x)Tix, what exactly the maximum solubility of Ti in face-centred cubic (fcc) Al is. Any change of symmetry for >2 at.% Ti beyond the known Fm3m Al98Ti2 phase [15] requires supporting structure factor calculations. As such, the nature of the strong variation in unit cell dimension across the 0–25 at.% Ti range in Ref. [3] must be clarified. The maximum solubility of Ti in Al in this concentration range is not rigorously proven, and the literature is split between two possible scenarios [3]. The first model claims that the maximum solubility is Al98Ti2, based on the early work on rapidly quenched melts in Ref. [15]. At compositions in between Al98Ti2 and Al75Ti25, the observed fcc phase never exceeds Al98Ti2, and the rest of the Ti is stored as an amorphous phase, or in an extended grain boundary network (in typical micron-dimensioned powder resulting from milling of Al–Ti mixtures). Based on extrapolation of the data in Ref. [15], some authors have claimed increased solubility to approximately Al96Ti4 [16], and even further to Al90Ti10 [17]. The second model proposes that in between Al98Ti2 and Al75Ti25, all of the Ti is present in the observed fcc phase, and it should be regarded as a ‘‘partially ordered” L12 structure type, discussed in Ref. [3] as a possibility for Ti concentrations >20 at.%, in the early stages of milling a 25 at.% mixture in Ref. [33], and for 10 at.% compositions in Ref. [17].

Cu Kβ

6

7

8

9

10

Fig. 11. EDS from the Al(1x)Tix crystallite in Fig. 10a (solid line) compared with a spectrum (dotted line) from a larger, average part of the sample. The two spectra are normalized with respect to the Al K edge.

3.2.1. Structure factor analysis of literature data for Al(1x)Tix and Al(1x)Zrx x 6 0.25 phases We shall describe in the following that features of both models 1 and 2 above are evident in our own data, and in older diffraction data from the Al–Ti and Al–Zr literature.

4698

M.P. Pitt et al. / Acta Materialia 56 (2008) 4691–4701

In the study of Al(1x)Tix sputter deposits with up to 49 at.% Ti [34], weak 1 0 0 and 1 0 1 reflections of the L12 primitive unit cell are observed in a multiphase Al90.5Ti9.5 sample. The sample contains a mixture of pure Al and an ˚ clearly indiL12 phase. The unit cell dimension of 3.9667 A cates a fully ordered L12 Al75Ti25 composition; however, the intensity ratio of 1 1 1/1 0 0 is 30, much larger than the normal ratio of 11.0 in the fully ordered structure. This indicates that Ti and Al have exchanged across their sublattices, yielding strongly quenched primitive reflections, as observed in the data (the ratio of 1 1 1/2 0 0 is 2.1, as expected, and is not affected by any possible preferred orientation, thus the 1 1 1/1 0 0 ratio of 30 is genuine). Using the X-ray scattering factors from the International Tables [35] (the same as used in RIETICA), calculation of the structure factors for the intensity ratio I111/I100 = 30 indicates that 8.8% of the Al sublattice must be occupied by Ti atoms, commensurate with 26.4% of the Ti sublattice being occupied by Al if the 3:1 stoichiometry is preserved (Al is typically solved into a Ti up to 45 at.% [36]), giving a final element distribution of (Al0.912Ti0.088)3Al0.264Ti0.736. Such sublattice mixing is a commonly observed feature when doping the L12 structure [10]. L12 primitive reflections are not typically observed in ball-milled Al(1x)Tix powders at intermediate 0 < x < 25 at.% Ti concentrations; however, the mechanical alloying study of Al-10 at.% X (X = Ti, Zr, Hf) alloys in Ref. [13] shows a single-phase Al90Zr10 X-ray diffraction pattern with very intense 1 0 0 and 1 1 0 reflections of the L12 primitive unit cell. Ti and Zr reside in the same group of the periodic table, and it is stated in Ref. [13] that both transition metals form a similar ‘‘partially ordered” L12 structure type in the 0–25 at.% X range. Extracting the intensities from the Al90Zr10 X-ray diffraction pattern in Ref. [13] gives a factor of 7.6 between the 1 0 0 and 1 1 1 intensities, a factor of 3.0 between 2 0 0 and 1 0 0, and a factor of 1.4 between 1 0 0 and 2 2 0. The

1 1 1/2 0 0 ratio is 2.5, higher than the expected value of 2.1 (invariant with sublattice swapping and Al(1x)Zrx elemental distribution); thus there is preferred orientation present on 1 1 1 of this data. As such, model intensity ratio calculations used below utilize the 2 0 0/1 0 0 and 2 2 0/1 0 0 ratios. Fig. 12a shows in the upper diagram the calculated diffraction pattern from the fully ordered stoichiometric L12 Al75Zr25 phase. The intensity ratio between 1 0 0 and 1 1 1 is 2.2 for Al75Zr25, considerably smaller than that which is observed experimentally for the Al90Zr10 phase in Ref. [13], 7.6 (6.3 when 1 1 1 is corrected for preferred orientation). The lower diagram in Fig. 12a is a reproduction (at X-ray synchrotron resolution) of the Al90Zr10 Xray diffraction in Ref. [13] with 6.4 1 1 1/1 0 0 ratio, and 1.8 2 2 0/1 0 0 ratio. A strong reduction in 1 0 0 and 1 1 0 intensity is clearly observable in the lower diagram of Fig. 12a. This reduction has been achieved by replacing 50% of the Zr sublattice with Al atoms. Attempts to model a large percentage of Zr atoms on the Al sublattice resulted in almost completely extinguished 1 0 0 and 1 1 0 intensity, and only 1.6% of Al atoms were replaced to achieve the closest factors of 6.4 and 1.8 described above. This amount of sublattice exchange yields an Al86.3Zr13.7 stoichiometry. This model can be further refined with structure factor calculations, and on calculating I220/I100 for the exact Al90Zr10 composition, only 0.5 at.% of Zr exists on the Al sublattice, and the final elemental distribution is Al90Zr10 = (Al0.995Zr0.005)3Al0.615Zr0.385. It is clear that such a mixed sublattice model could be refined further with new high-resolution X-ray synchrotron data, and the Al(1x)Zrx system is clearly an ideal system to test as strong primitive 1 0 0 and 1 1 0 reflections are clearly observable, at least for the Al90Zr10 phase. As the 1 0 0 intensity is very sensitive to the amount of Zr present on the Al sublattice, it is highly desirable also to follow the system as a function of milling time, to observe if the Al sublattice accepts more

L12 (311)

L12 (220)

(300) (310)

L12 (200)

Al75Ti25

L12 (210) L12 (211)

L12 (110)

L12 (100)

3000 Intensity (counts)

L12 (311)

L12 (220)

(300) (310)

4000

L12 (200) L12 (210) L12 (211)

(111) L12 (110)

Intensity (counts)

L12 (100)

6000

(111)

4000 Al75Zr25

2000

2000 Al90Zr10

Al85Ti15

1000

0 5

10

15 20 2θ (degrees)

25

5

10

15 20 2θ (degrees)

25

Fig. 12. (a) Fully ordered L12 Al75Zr25 (upper figure), and a reproduction of the Al90Zr10 pattern observed in [13] (lower figure). (b) Fully ordered L12 Al75Ti25 (upper figure), and a mixed sublattice Al85Ti15 L12 pattern with quenched 1 0 0 and 1 1 0, etc., primitive reflections.

M.P. Pitt et al. / Acta Materialia 56 (2008) 4691–4701

Zr over time, which would result in almost extinguished 1 0 0 and 1 1 0 reflections. The Al90Zr10 system in Ref. [13] has been milled for 20 h in a SPEX8000 shaker mill. 3.2.2. Quenching of the L12 superlattice for Al(1x)Tix x60.25 phases Fig. 12b shows the stoichiometric fully ordered L12 Al75Ti25 diffraction pattern (upper figure). 1 0 0 and 1 1 0 reflections of the primitive unit cell are clearly observable. Starting with the same amount of sublattice swapping for the Al90Zr10 model above, it was clear that for the Ti system, very weak intensity remained in 1 0 0 and 1 1 0. In the same manner as described above for the Zr system, more Ti atoms were placed on the Al sublattice, which resulted in very strong quenching of 1 0 0 and 1 1 0. It thus becomes prudent to examine the necessary crystallographic conditions to quench to zero the primitive reflections of partially ordered 0 < x < 25 at.% L12 Al(1x)Tix structure types. Maintaining the 3:1 ratio of the Al:Ti sublattices requires a mixed (Al1xTix)3(AlyTi1y) structure type. The relevant stoichiometry, e.g. Al85Ti15 = Al3.4Ti0.6, is then attained by forcing 3(1x) + y = 3.4, etc. Calculation of the structure factor about 1 0 0 and subsequently forcing it to zero yields simultaneous equations, which when solved demonstrate that the original Al:Ti ratio is forced through both sublattices as e.g. Al85Ti15 = (Al0.85Ti0.15)3(Al0.85Ti0.15). Exchange of atoms in varying ratios across each sublattice of the L12 structure can clearly produce any composition in the 0–25 at.% Ti range, and the presence of L12 superlattice reflections is sensitive to the amount of sublattice exchange. Equal mixing across both sublattices clearly produces an A1 type structure. The lower diagram in Fig. 12b shows the L12 structure with 15% of the Al sublattice replaced with Ti atoms, and 85% of the Ti sublattice replaced with Al atoms, giving an Al85Ti15 composition. It is clear that the lower diffraction pattern in Fig. 12b appears as a normal fcc Fm3m type pattern, with no clear indication of the primitive cell. This is the type of Al(1x)Tix pattern that is consistently observed in our own data, and in ball-milled samples in the Al–Ti literature. Depending on the amount of Ti substitution on the Al sublattice, it is possible for weak L12 reflections to remain. Such weak primitive reflections (tens of counts) may be lost in background noise (several hundred to 1000 counts), even in data with intense Bragg reflections (50,000 counts). This problem is further compounded by the strong size/strain broadening observed in our own work (heavily h1 1 0i{1 1 1} edge-dislocated c-Al(1x)Tix phases (c = crystalline) of very small dimension, 4–25 nm), and also in the Al–Ti literature [3,37], which would result in weak broadened primitive reflections that are difficult to observe in background noise. 4. Discussion It should be emphasized that our study focuses on the properties of isolated nanocrystallites of Al(1x)Tix x < 0.25. The Al–Ti literature has in general focused on

4699

micron-dimensioned Al(1x)Tix x60.25 phases. However, it can be assumed that when mixing elemental Al and Ti powders (typically by ball milling), the properties observed from our study will be valid in the folded nanoscale mixing regions between Al and Ti. We present examples in the discussion below. When ball milling micron-dimensioned elemental powders of Al and Ti, the unit cell of the observed fcc Al(1x)Tix phase continues to contract across the entire Al98Ti2 to Al75Ti25 range. Ti apparently continues to enter the Al lattice; however, there is no high-resolution diffraction work that follows the simultaneous behaviour of both the Al and Ti unit cells before the Ti disappears. In the diffraction patterns in Ref. [16], Ti does not completely disappear in a 25 at.% Ti mixture milled for 40 h, with broad peaks remaining at the Ti positions, indicating fine residual Ti, probably nanocrystals. The mechanics of the Al/Ti interface (nanocrystals of Ti immersed in a larger Al(1x)Tix matrix) may produce considerable strain due to partial coherency at the interface, and the Al(1x)Tix unit cell dimension may be locally affected; however, a large homogeneous strain to contract the cell from pure Al dimensions to close to Al75Ti25 dimensions is extremely unlikely. However, in Ref. [37], a strong correlation is evident between the Al(1x)Tix unit cell dimension and its mosaic size. We also observe similar features in our data, with Al80Ti20 appearing to be stabilized by a high dislocation density in long-term PM NaAlH4 + xTiCl3. In Ref. [3], it is observed that diffraction peaks of milled Al(1x)Tix samples become weaker and broader with increasing Ti concentration. For the same amount of broadening, the continual addition of Ti atoms to the fcc Al phase will increase the X-ray scattering intensity as a function of Ti concentration. With increasing broadening, this indicates that another Ti-containing phase must be present to ‘‘weaken” the diffraction reflections. The coexistence of amorphous Al(1x)Tix phases with the fcc Al(1x)Tix solid solution is reported in melt spins up to Al97Ti3 to Al88Ti12 in Ref. [38]. Such evidence favours model 1 (maximum crystalline content Al98Ti2, remaining Ti in amorphous state); however, a metallurgical explanation for the contraction of the fcc unit cell with increasing Ti content is not clear. The presence of fine amorphous Al(1x)Tix phase is clear from the SEM work in Ref. [39] that shows a lamellae microstructure is initially formed between Al and Ti, which can form fine interfacial amorphous Al(1x)Tix phases. Such morphology is not uncommon for amorphous phases within a crystalline matrix, and has even been utilized on a macroscopic scale for bulk composites consisting of two metal sheets co-deformed by mechanical working [40], with amorphous phases occurring at the sub-micron interface by thermally assisted diffusion from each sheet. These features are also commonly observed by the folding process inherent to mechanical ball milling [39]. For the NaAlH4 + xTiCl3 system, a-Al(1x)Tix phases (a = amorphous) are formed at rich TiCl3 content (>5%) by the highly localized reduction and formation of metallic Ti and Al, resulting

4700

M.P. Pitt et al. / Acta Materialia 56 (2008) 4691–4701

in extremely fine (2–20 nm), well-dispersed and isolated amorphous Al(1x)Tix (0.3 < x < 0.5) phases [22]. After one of the three treatments described in Section 3.1 above, Ti diffuses out of the amorphous Al(1x)Tix (0.3 < x < 0.5) matrix and forms crystalline Al(1x)Tix x < 0.25 phases. Depending on the duration of the treatment, quantitative phase analysis indicates that a certain fraction of Ti remains bound in the amorphous matrix, realizing a localized nanoscopic mixture of amorphous and crystalline a-Al(1x)Tix + c-Al(1x)Tix phases, with x0.15. This is similar to the reported coexistence of amorphous Al(1x)Tix phases with the fcc Al(1x)Tix solid solution in Ref. [38]. The correct crystal structure for crystalline Al(1x)Tix x < 0.25 phases can be chosen from either the A1 or L12 models, and is dependent on the observed intensity of superlattice reflections of the L12 unit cell, such as 1 0 0. An extended Al(1x)Tix Fm 3m solid solution model is identical in intensity to a 1 0 0 quenched L12 Al(1x)Tix structure, and both models display the same amount of Ti substitution on the Al sublattice. The L12 model is strongly suggested from several points of view, as (i) Ti and Zr are from the same group of the periodic table, and we thus expect a similar chemical bonding environment; (ii) there is unambiguous experimental evidence showing clear and intense L12 reflections for single-phase Al90Zr10 [13]; (iii) exchange across sublattices is clearly indicated by the high 1 1 1/1 0 0 ratio of 30 for Al75Ti25 in Ref. [34]; and (iv) the difference in magnitude of X-ray scattering factors between Zr and Ti, in combination with the low diffusivity flux of Zr in Al [41] (the lowest of all transition metals) ensures the Al(1x)Zrx system will always display more intense primitive L12 reflections compared to the Al(1x)Tix system. Such understanding is clearly of general interest for both the TiCl3-enhanced NaAlH4 system, as well as the Al–Ti field. The authors of Ref. [13] also report L12 structure types for both Al90Ti10 and Al90Hf10, although the diffraction patterns are not presented. However, as discussed above, superlattice reflections such as 1 0 0 are very difficult to find in the Al–Ti literature, and in our own data. As their contribution to the data appears negligible, the use of the Fm 3m model is equally valid in our data, and in any case displays the same Ti substitution on the Al sublattice. Such features imply that the Al98Ti2 solubility limit [15] is necessarily extended, and either the A1 or L12 structure models are valid to use, depending on the presence of weak superlattice reflections. Differences in milling technique and milling time (20 h in a shaker type mill in Ref. [13] compared to 1 h planetary milling in our samples) may be sensitive to the development of the final intensities in the superlattice L12 reflections. Further, it is clear that depending on the elemental ratio, some compounds can display the disordered A1 structure ðFm 3mÞ, the ordered L12 structure ðPm 3mÞ or the ordered L10 structure (P4/ mmm) [42]. The transition between these structure types is reported to be sensitive to microstructure [43]. The high h1 1 0i{1 1 1} edge dislocation density of 6  1016 m2 in small 25 nm Al(1x)Tix crystallites is

not typically observed on the nanoscale in the Al–Ti literature, nor in elemental nanoscopic fcc metals. The quest to find densely dislocated nanograins to study the limits of plasticity has typically been pursued in subgrain nanomosaics, within powder grains that possess external dimensions on the microscale. As such, the well-dispersed and isolated nanocrystallites of Al85Ti15 in our samples present a new and unique case in which to study the presence of dislocations in isolated nanocrystallites. The initial reduction of NaAlH4 + xTiCl3 ? (1  3x)NaAlH4 + 3xNaCl + xTi + 3xAl + 6xH2(g) (Reaction 1) is similar in concept to the synthesis of pure nanoscopic metals by reduction of transition metal-salts against Na, such as NiCl2 + 2Na ? 2NaCl + Ni [44], resulting in a fine 10– 20 nm dispersion of isolated nanocrystallites of Ni on the NaCl surface. However, high dislocation densities are not typically observed in these isolated nanometal crystals, and further treatment such as cryomilling is typically necessary to dislocate the crystals. We note that the 2–20 nm Al formed by reduction (Reaction 1) during the milling process already contains a similar high density of h1 1 0i{1 1 1} edge dislocations to the 4–25 nm Al85Ti15 phase. That is, both the base crystal structure and defect structure are already formed during Reaction 1, and all that is required to form crystalline Al(1x)Tix x < 0.25 phases is the local diffusion of Ti atoms out of the amorphous Al(1x)Tix (0.3 < x < 0.5) matrix, into the defected Al nanocrystals. We speculate that the atypical high h1 1 0i{1 1 1} edge dislocation density in the 2–20 nm Al is a result of H-assisted nucleation, due to the local presence of a partial pressure of H2 (produced from Reaction 1) within the high-purity Ar atmosphere, in the immediate vicinity of the reduction site. We note also that the upper limit of the size range for the Al85Ti15 phase is very close to the theoretical crossover value from dislocation-mediated plasticity to grain boundary sliding of 18 nm for pure Al [45], and is typical of the 10–20 nm crossover size estimated for fcc metals [46]. Morphologically, it is not possible for the h1 1 0i{1 1 1} edge dislocations to glide (under the application of temperature) from one nano Al(1x)Tix grain to another and/or be absorbed into a grain boundary, as the Al(1x)Tix crystallites are isolated from each other, and well dispersed and embedded on the NaAlH4 surface. It is also likely that once Ti atoms enter the original defected 2–20 nm Al crystallites (which clearly occurs at a temperature lower than the dislocation annealing temperature), the Ti atoms can also then act as pinning centres, preventing any temperature-dependent dislocation annihilation. This is commensurate with the observation that the Al85Ti15 peak shape appears the same after 100 H cycles [25]. It is clearly desirable to study with TEM the neighbouring ScCl3 and VCl3-enhanced NaAlH4 systems, to observe if a similar dislocation density occurs in the nanoscopic Al(1x)Scx or Al(1x)Vx phases that are formed, as these phases also possess the same related base symmetry and structure type (Al: Fm3m; Al(1x)Tix: Fm 3m; Al3Ti: Pm3m; Al3Sc: Pm3m; Al3 V: I4/mmm).

M.P. Pitt et al. / Acta Materialia 56 (2008) 4691–4701

5. Conclusion Mixed L12 sublattice structure factor calculations for both the Ti and Zr system agree with observed experimental intensities in X-ray diffraction data, and together with local EDS measurements, Ti is observed to be continuously solved into the Al sublattice, at least up to Al85Ti15. Nanoscopic crystalline Al(1x)Tix (x < 0.25) phases can be observed coexisting with nanoscopic amorphous Al(1x)Tix (x  0.15) phases for rich TiCl3 content. For our data, model 1 (maximum crystalline content Al98Ti2, remaining Ti in amorphous state) and model 2 (all Ti stored in crystalline, partially ordered L12 phase) can clearly exist simultaneously, and as such, for ball-milled bulk microndimensioned mixtures of Al and Ti powders, we expect that Ti will be continuously solved into Al as c-Al(1x)Tix + a-Al(1x)Tix up to 25 at.% in nanoscale (<25 nm) mixing regions between pure Al and Ti. Crystalline Al(1x)Tix (x < 0.25) phases observed in NaAlH4 + xTiCl3 samples are most appropriately described as L12 superlatticequenched, A1 equivalent structure types. The correct determination of crystal structure for Al85Ti15 also serves to accurately quantify the amount of Ti present in c-Al(1x) Tix (x < 0.25) phases in the NaAlH4 + xTiCl3 system. Acknowledgements M.P.P. acknowledges support from the Synchrotron Program of the Norwegian Research Council, and also thanks the staff of the SNBL for their skilful assistance during long-term attachments.

[10] [11] [12] [13]

[14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27]

[28] [29] [30] [31] [32] [33] [34] [35]

[36]

References [1] Froes FH, Suryanarayana C, Eliezer D. J Mater Sci 1992;27:5113. [2] Milman YuV, Miracle DB, Chugunova SI, Voskoboinik IV, Korzhova NP, Legkaya TN, et al. Intermetallics 2001;9:839. [3] Schoenitz M, Zhu X, Dreizin EL. J Met Nanocryst Mat 2004;2021:455. [4] Yamaguchi M, Umakoshi Y. Prog Mater Sci 1990;34:1. [5] Zhang S, Nic JP, Mikkola DE. Scripta Metal 1990;24:1099. [6] Nayak SS, Murty BS. Mat Sci Eng A 2004;367:218. [7] Chaki TK. Phil Mag Lett 1990;61:5. [8] Chaki TK. Phil Mag Lett 1990;63:123. [9] Rossouw CJ, Forwood CT, Gibson MA, Miller PR. Phil Mag 1996;A74:57.

[37] [38] [39] [40] [41]

[42] [43] [44] [45] [46]

4701

Rossouw CJ, Forwood CT, Gibson MA. Phil Mag 2000;A80:1583. Forwood CT, Gibson MA. Mater Sci Forum 1990;189-190:353. Watanabe Y, Eryu H, Matsuura K. Acta Mater 2001;49:775. Schwarz RB, Desch PB, Srinivasan SR. In: deBarbadillo JJ, Fores FH, Schwarz R, editors. Proceedings of the 2nd international conference on structural applications of mechanical alloying. Materials Park, OH: ASM; 1993. p. 227. Schuster JC, Palm M. J Ph Eq Diff 2006;27:255. Tonejc A, Bonefacic A. Scripta Metal 1969;3:145. Zhang F, Lu L, Lai MO. J Alloys Comp 2000;297:211. Fan GJ, Quan MX, Hu ZQ. J Mater Sci 1995;30:4847. Draissia M, Debili M-Y. Cen Euro J Phys 2005;3:395. Axon HJ, Hume-Rothery W. Proc R Soc Lond 1948;193A:1. Bogdonavic B, Schwickardi M. J Alloys Comp 1997;253-254:1. Hunter BA. Comm Powder Diffraction Newsl 1998;20:21. Pitt MP, Vullum PE, Sørby MH, Blanchard D, Sulic MP, Walmsley JC, et al. Submitted for publication. Majzoub EH, Gross K. J Alloys Comp 2003;356-357:363. King HW. J Mater Sci 1966;1:79. Brinks HW, Hauback BC, Srinivasan SS, Jensen CM. J Phys Chem B 2005;109:15780. Herley PJ, Jones W. Mater Lett 1983;131:1. Vullum PE, Pitt MP, Walmsley JC, Hauback BC, Holmestad R. Transmission electron microscopy characterisation of NaAlH4, submitted for J Phys Chem C. Wu E, Gray EMacA, Kisi EH. J Appl Crystallogr 1998;31:356. Wu E, Kisi EH, Gray EMacA. J Appl Crystallogr 1998;31:363. Krivoglaz MA, Martynenko OV, Ryaboshapka KP. Fiz Met Metallov 1983;55:5. Wilkens M. Phys Status Sol A 1970;2:359. Williams DB, Carter CB. Transmission electron microscopy, vol. IV. New York: Plenum Press; 1996. Chap. 35. Klassen T, Oehring M, Bormann R. J Mater Res 1994;9:47. Yoshioka H, Yan Q, Asami K, Hashimoto K. Mat Sci Eng 1991;A134:1054. Malsen EN, Fox AG, O’Keefe MA. In: Wilson AJC, editor. International tables for crystallography, vol. C. Dordrecht: Kluwer Academic; 1992. p. 476. Massalski TB, editor. Binary alloy phase diagrams handbook, Al–Ti system. Materials Park, OH: ASM; 1990. Zhang F, Lu L, Lai MO, Froes FH. J Mater Sci 2003;38:613. Kimura H, Sasamori K, Inoue A. Mat Trans JIM 1998;39:773. Oehring M, Klassen T, Bormann R. J Mater Res 1993;8:2819. Atzmon M, Unruh K, Johnson WL. J Appl Phys 1985;58:3865. Adam CM. In: Kear BH, Giessen BC, Cohen M, editors. Rapidly solidified amorphous and crystalline alloys. New York: Elsevier; 1982. p. 411. Dai ZR, Sun S, Wang ZL. Nano Lett 2001;1:443. Pankratov IR, Vaks VG. J Phys Condens Matter 2001;13:6031. Tsuzuki T, McCormick PG. J Mater Sci 2004;39:5143. Yamakov V, Wolf D, Phillpot SR, Mukherjee AK, Gleiter H. Phil Mag 2003;A83:385. Li H, Fereshteh E. Acta Mater 2006;54:2877.