Microstructural Investigation of Nanocrystalline Hydrogen-Storing Mg-Titanate Nanotube Composites Processed by High-Pressure Torsion

A high-energy ball milling and subsequent high-pressure torsion method was applied to synthesize nanocrystalline magnesium samples catalyzed by TiO2 or titanate nanotubes. The microstructure of the as-milled powders and the torqued bulk disks was characterized by X-ray diffraction. The recorded diffractograms have been evaluated by the convolutional multiple whole profile fitting algorithm, which provided microstructural parameters (average crystal size, crystallite size distribution, average dislocation density). The morphology of the nanotube-containing disks has been examined by high-resolution transmission electron microscopy. The effect of the different additives and preparation conditions on the hydrogen absorption behavior was investigated in a Sieverts’-type apparatus. It was found that the ball-milling route has a prominent effect on the dispersion and morphology of the titanate nanotubes, and the absorption capability of the Mg-based composite is highly dependent on these features.


Introduction
The coupling between continuing growth of the world's population and industrialization means that the transportation segment of the economy is responsible for significant part of petroleum consumption, and accordingly transportation is one of the main source of global CO 2 emission [1]. In addition, the world's energy consumption is expected to be doubled by 2050, which requires the introduction of alternative energy sources to the energy grid [2]. Even though these alternative renewable sources are not competitive with fossil fuels at the moment [3], hydrogen as a secondary energy carrier has gained increasing attention in the last decade; however, significant technical and economical challenges, such as efficient production and storage need to be solved [4]. It has been reported that hydrogen can be stored either in molecular form by physisorption on the surface of carbon-based materials [5,6], or in atomic form by chemisorption in metal-based compounds [7,8].
As has been extensively studied, absorption of hydrogen in metal-based compounds results in higher energy density and safety when compared to gas storage of molecular hydrogen or

Preparation of Titanate Nanotubes
Titanate nanotubes (TN) were prepared via the microwave-assisted hydrothermal method [47,48]. The starting material was anatase TiO 2 powder (supplied by Reanal; purity 99.9%) which was filtered using cotton wool and distilled water. 0.35 g of filtered TiO 2 powder was mixed with 7 mL of NaOH aqueous solution (10 M), the suspension was stirred ultrasonically for 20 min. The mixture was then heated to 135 • C for 8 h in a closed SiC vial by microwave irradiation using a Monowave 300 microwave reactor in the presence of continuous stirring. The product was then dispersed in 300 mL distilled water and centrifuged for 5 min at 1500 RPM, this procedure was repeated two more times. The remaining material was dried overnight at 60 • C. As seen in Figure 1, the as-prepared titanate nanotubes are multi-wall tubes, with a length ranging from 40 nm to several hundreds of nanometers and their diameter is in the range of 10 nm.

Preparation of Magnesium-Titanate Nanotube Composites
Mg-TN composites were prepared by mixing Mg powder (supplied by Sigma-Aldrich; purity 99.9%, initial powder particle size~100 µm), 5 wt.% titanate nanotube (described in the previous section) and 3 mL n-hexane. The resulting suspension was ball-milled in a stainless steel vial (volume: 65 mL) using a SPEX 8000M Mixer Mill operating at 1425 RPM. For all samples, 10 stainless steel balls (1/4 in.) were used, thus the ball-to-powder weight ratio was 10:1. Two milling routes were applied: all the constituents were milled together for 3 h (denoted as MgTN_3h); 2.
the magnesium powder was pre-milled for 2.5 h in hexane before the titanate nanotubes were added for an additional 0.5 h milling (denoted as MgTN_0.5h).
For comparison, a Mg-TiO 2 composite was also prepared, using 5 wt.% TiO 2 commercial powder via wet ball-milling in 3 mL n-hexane (denoted as MgTiO2_3h). Here, milling route #1 was applied, i.e., the materials were milled together for 3 h, other milling parameters were the same as above. Sample handling was carried out in a glove box and HEBM was processed in argon atmosphere in order to minimize the oxidation of the powder samples.
The as-milled powders were pre-compacted into pellets with a diameter of 8 mm and height of 0.8 mm under~200 MPa of hydraulic pressure. High-pressure torsion of these pellets was carried out in air, under an applied pressure of 4 GPa at room temperature. N = 3 and N = 10 whole revolutions were performed for all types of powder samples with an angular speed of ω = 0.2 revolution/min. These disk shape samples will be denoted as: MgTN_3h_N3, MgTN_3h_N10, MgTN_0.5h_N3, MgTN_0.5h_N10, MgTiO2_3h_N3 and MgTiO2_3h_N10.

Microstructural Characterization
The overall microstructure of the powders and HPT-disks was examined by X-ray powder diffraction (XRD). The measurements were carried out on a Rigaku SmartLab diffractometer in θ-2θ geometry using Cu-Kα radiation. The data were collected from 20 • to 75 • with a step size of 0.01 • . The recorded powder diffraction patterns have been evaluated by the Williamson-Hall method [49]. The average crystallite size (D WH ) and the square root of the mean square strain ( ε 2 1/2 ) can be determined from the positions and widths of the Bragg reflections according the following equation [49]: where K = 2sinθ/λ and ∆K = 2cosθ(∆θ)/λ. Here θ is the Bragg angle, ∆θ denotes the half of the full width at half maximum of the crystalline peaks, A is geometrical constant, usually taken as A = 1 and λ is the wavelength of the radiation (λ = 0.154 nm). In order to reveal the possible differences in microstructure along the radius of the HPT disks, a Rigaku rotating anode high resolution X-ray diffraction instrument with Cu-Kα radiation was also applied. The spot size of the X-ray beam was 1 mm x 0.1 mm which enabled us to characterize separate regions with different ultimate strain along the disks, i.e., center, half-radius and edge regions were studied. The operating voltage and current were 40 kV and 80 mA, respectively. For detection of the diffracted X-ray photons, image plates (IP) were used at a distance of 200 mm from the sample.
The XRD profiles measured by the rotating anode instrument have been evaluated by the convolutional multiple whole profile (CMWP) fitting algorithm which incorporates the direct fit of the whole measured diffraction profile by the sum of theoretically constructed profile functions, background and measured instrumental profiles [50]. In brief, the contribution of lattice strain and coherently scattering domains on peak broadening can be separated on the basis of their different diffraction order dependence. The applied profile functions are generated for each Bragg-reflection of each crystalline phase as the inverse Fourier transform of the product of the size and strain Fourier coefficients providing both the strain and size parameters of the microstructure [51]. In this model, it is assumed that the crystallites have a lognormal size distribution: where σ and m are the variance and median of the histogram, respectively. As a result, the average size of the coherently scattering domains can be given as The CMWP algorithm also provides the strain parameters, i.e., the average dislocation density (ρ) can also be determined [50].

Transmission Electron Microscopy (TEM)
Local structure of the HPT-disks were characterized by high-resolution transmission electron microscopy (HR-TEM) using a JEOL JEM 3010 electron microscope equipped with a LaB 6 cathode and working at 300 keV (point resolution 1.7 Å). Sample preparation for TEM investigations was carried out with focused ion beam (FIB) milling, i.e., a lamella was cut out from the disk (close to the edge) and subsequently thinned in a FEI QUANTA 3D scanning electron microscope equipped with FIB.

Hydrogen Storage Experiments
The hydrogen storage properties of the samples were investigated using a home-made Sieverts' type apparatus. Isothermal absorption and desorption measurements were performed at 573 K, with an initial hydrogen pressure of 1 MPa and 1 kPa, respectively. The actual measurements were carried out after two full hydrogenation-dehydrogenation cycles. For all the experiments, the whole HPT-disks (~80-90 mg each) were used as prepared.

Results and Discussion
The XRD patterns of the as-milled composites along with the commercial Mg and TiO 2 powders and the as-prepared titanate nanotubes are presented in Figure 2. As is apparent, the main constituent phase is hexagonal Mg (JCPDS: 00-035-0821) for all composite samples. In addition, Bragg peaks of TiO 2 (JCPDS: 01-084-1285) can also be recognized for the MgTiO2_3h powder. On the other hand, no reflection corresponding to titanate nanotubes can be visualized in either nanotube containing samples (MgTN_0.5h and MgTN_3h), which might be attributed to the low crystallinity of the additive or to a possible partial amorphization, which can occur during the HEBM process when the milling balls impact the tubes. At the same time, a broad peak of MgO (JCPDS: 45-0946) at around 2θ = 43 deg is present for all the as-milled powders. Based on preliminary experiments, we chose two rotation numbers N = 3 and N = 10 for further processing the compacted pellets by HPT. The corresponding powder diffractograms averaging the X-ray scattering from the whole disks indicate only a slight deviation compared to those of the as-milled powders (Figure 3), i.e., all patterns are dominated by the crystalline peaks of hexagonal Mg. It is also noted that the intensity of the MgO phase remains practically the same after the HPT-processing carried out under air. As one can also notice, the peaks of TiO 2 are still visible after the HPT deformation (MgTiO2_3h_N3 and MgTiO2_3h_N10 samples). The relative intensities of the Mg peaks vary considerably between specimens prepared by the different routes (see Table 1). Notably, the intensity ratio of the (101) and (002) Mg reflections can be used as a measure of texture generated during deformation processes [40]. As one can recognize, the HEBM procedure results in some reduction of the I(101)/I(002) ratio, as compared to the commercial Mg powder (initially I(101)/I(002) = 3.04). However, a more prominent change can be observed as a result of the torsion process, i.e., I(101)/I(002) decreases down to the 0.3-0.4 interval even after N = 3 whole revolutions. Interestingly, additional shear deformation (N = 10) does not further alter the texture of the Mg phase for the MgTN_0.5h type sample, meanwhile the I(101)/I(002) ratio slightly increase and decrease for the MgTN_3h and MgTiO2_3h type samples, respectively. The average grain size values (D WH ) and the square root of the mean square strain ( ε 2 1/2 ) for all the as-milled powders and HPT-disks were obtained by the Williamson-Hall analysis and are plotted in Figure 4a,b, respectively. As can be seen, some variation takes place in the D WH values (65-85 nm range) of hexagonal magnesium when it is milled together with different additives. The HPT-deformation results in some further nanocrystallization, i.e., the decrease of the average grain size for the TiO 2 containing composite (MgTiO2_3h) and for the sample MgTN_3h occurs between N = 3 and N = 10 revolutions (D WH~5 5-60 nm). A similar reduction can be observed in case of the MgTN_0.5h composite, however, this change occurs already during the first 3 revolutions. Rather similar trends can be visualized for the mean square strain (see Figure 4b). As the powders have been subjected to HEBM, a ε 2 1/2~0 .37%-0.45% strain already develops as a result of the severe plastic deformation generated by the colliding balls. As HPT is introduced, a general decrease for all the composites can be recognized after N = 10 revolutions, reaching an ultimate value of ε 2 1/2~0 .35%, corresponding to some kind of strain release during the shear deformation. It is also noted that a slight increase can be observed of the MgTN_3h_N3 disk. As the HPT procedure imposes inhomogeneous strain along the radius of the disk, smaller revolution numbers can result in a deformation-dependent structure [52] and mechanical properties [53], however, larger rotations can homogenize the material [54,55]. Therefore, we chose the disks that performed for N = 3 rotations for subsequent investigations.
In order to explore the radius dependent microstructure, position sensitive high resolution XRD was carried out on different parts of the disks. Figure 5 shows one set of diffractograms taken at different positions of the MgTN_3h_N3 disk. As seen, the XRD profiles are qualitatively similar, i.e., hexagonal Mg reflections dominate the patterns. These high-resolution XRD profiles were analyzed by the CMWP method in order to obtain several microstructural parameters. A typical example of the measured and fitted profiles together with the difference plot corresponding to the half radius of the MgTN_3h_N3 sample unambiguously confirms that the numerical algorithm has satisfactorily converged (see Figure 6). Based on the evaluation of the line profiles by the CMWP method, the average coherently scattering crystallite size of Mg phase can be determined as shown in Figure 7a. In general, the D CMWP values are fairly constant along the radius for all the disks after N = 3 revolutions. Nevertheless, the type of additive has an influence on the crystallite size, i.e., the addition of titanate nanotubes results in smaller D CMWP of Mg than in case of TiO 2 catalyst, in accordance with Figure 4a. In spite of the apparent invariance of the average crystallite size along the disk radius, some microstructural differences between the different sectors can be identified. Using the m and σ parameters determined from the fitting procedure, the G(x) lognormal size-distribution functions can be generated (Figure 7b). In the case of the MgTN_3h_N3 disk, the histograms are markedly different for the edge region and for the rest of the disk, i.e., the distribution has smaller median, but larger variance corresponding to a more homogeneous nanostructure at the perimeter of the specimen, despite the rather similar average crystallite size values. Similar tendency was observed for the other two disks processed for N = 3 turns. Figure 7c presents the dislocation density values across the disk radius after N = 3 revolutions. For the MgTiO2_3h_N3 and the MgTN_3h_N3 samples, the ρ values are practically constant throughout the disk (1.7·10 15 m −2 and 2.1·10 15 m −2 respectively) which clearly indicates that the dynamic equilibrium between dislocation generation and annihilation is already present even after N = 3 rotations. Slightly different behavior can be seen for the MgTN_3h_N3 specimen, i.e., the center of the disk exhibits somewhat lower dislocation density than the rest of the disk.  Based on the evaluation of the line profiles by the CMWP method, the average coherently scattering crystallite size of Mg phase can be determined as shown in Figure 7a. In general, the DCMWP values are fairly constant along the radius for all the disks after N = 3 revolutions. Nevertheless, the type of additive has an influence on the crystallite size, i.e., the addition of titanate nanotubes results in smaller DCMWP of Mg than in case of TiO2 catalyst, in accordance with Figure 4a. In spite of the apparent invariance of the average crystallite size along the disk radius, some microstructural differences between the different sectors can be identified. Using the m and σ parameters determined from the fitting procedure, the G(x) lognormal size-distribution functions can be generated ( Figure  7b). In the case of the MgTN_3h_N3 disk, the histograms are markedly different for the edge region and for the rest of the disk, i.e., the distribution has smaller median, but larger variance corresponding to a more homogeneous nanostructure at the perimeter of the specimen, despite the rather similar average crystallite size values. Similar tendency was observed for the other two disks processed for N = 3 turns. Figure 7c presents the dislocation density values across the disk radius after N = 3 revolutions. For the MgTiO2_3h_N3 and the MgTN_3h_N3 samples, the ρ values are practically constant throughout the disk (1.7·10 15 m −2 and 2.1·10 15 m −2 respectively) which clearly indicates that the dynamic equilibrium between dislocation generation and annihilation is already present even after N = 3 rotations. Slightly different behavior can be seen for the MgTN_3h_N3 specimen, i.e., the center of the disk exhibits somewhat lower dislocation density than the rest of the disk. For local structural and morphological analysis, transmission electron microscopy was carried out (Figure 8). The low magnification micrograph of the MgTN_0.5h_N3 sample presents inhomogenities in the HPT-disk as several darker regions can be recognized on the lamella ( Figure  8a). An enlarged view, Figure 8b, shows one of these areas where straight bundles of nanotubes can be recognized along with some amorphous-like regions. Noticeably different features can be For local structural and morphological analysis, transmission electron microscopy was carried out (Figure 8). The low magnification micrograph of the MgTN_0.5h_N3 sample presents inhomogenities in the HPT-disk as several darker regions can be recognized on the lamella (Figure 8a). An enlarged view, Figure 8b, shows one of these areas where straight bundles of nanotubes can be recognized along with some amorphous-like regions. Noticeably different features can be identified for the MgTN_3h_N3 disk (Figure 8c,d). On the lower magnification image similar inhomogenities can be seen as for the MgTN_0.5h_N3 disk (Figure 8c), however, these inhomogenities are finer and their density is significantly higher compared to Figure 8a. The better distribution of the nanotube additive material is not surprising due to the longer milling time. Closer examination again proves that the darker areas contain titanate nanotubes (see Figure 8d), nevertheless the 3 h of milling partially destroys the tube structure as significantly amorphized regions can be identified in the vicinity of the tubes. It is noteworthy that the remaining nanotubes are considerably shorter and more curved than the fairly straight and mostly intact nanotubes of MgTN_0.5h_N3 sample.   Figure 7c). However, when the titanate nanotubes are milled together with the Mg powder for longer durations (MgTN_3h_N3 disk), the observed kinetics and maximum storage capacity (4.7 wt.%) seemingly exceed those of the TiO 2 catalyzed sample, which means that properly dispersed TNs can effectively enhance the hydrogen storage properties of Mg. The significant differences obtained between the storage properties of the MgTN_0.5h_N3 and MgTN_3h_N3 disks cannot be solely related to the microstructural parameters of the hexagonal magnesium phase obtained from the CMWP analysis, since they exhibit very similar D CMWP values and comparable dislocation density throughout their radii (see Figure 7a,c). Likewise, the detected minor variation in the I(101)/I(002) texture cannot be responsible for such distinct hydrogenation performance [40]. On the other hand, the TEM micrographs present substantial differences in the morphology and distribution of the titanate nanotube additives (see Figure 8). Therefore, we can conclude that the better dispersion of TN catalyst after 3 h of milling together with the Mg powder has a vital effect on the hydrogen uptake capability of the material. The defected nature of the nanotubes in the MgTN_3h_N3 disk could also be accounted for enhanced H-storage properties, due to the higher activity of destructed TN regions towards hydrogen [56]. Table 2 presents some other Mg-based hydrogen storage materials processed by HPT for comparison. As seen, the addition of titanate nanotubes to Mg by the combination of HEBM and HPT of the present work (MgTN_3h_N3) results in a higher hydrogen storage capability than that of pure Mg processed by HPT [57,58]. Nevertheless, the achieved capacity is somewhat lower than was obtained for Mg catalyzed by Nb 2 O 5 powder and carbon nanotubes synthesized under similar deformation conditions [59], and therefore the extension of this investigation is required. The obtained storage capacity of our MgTN_3h_N3 disk is similar to the value of magnesium milled with 5 wt.% TiO 2 powder for 10 h [60]. Very recent works on ball-milled nanocrystalline MgH 2 catalyzed by special flower-like TiO 2 [61] or TiO 2 nanosheet [62] additives resulted in a somewhat better hydrogenation performance.

Conclusions
The HEBM + HPT deformation route was applied to Mg powders catalyzed by titanate nanotubes (MgTN_0.5h_N3 and MgTN_3h_N3) and TiO 2 (MgTiO2_3h_N3). It was shown that HPT processing can further decrease the average crystallite size of ball-milled catalyzed Mg (from 65-85 nm to 55-60 nm) and can also introduce texture. In order to study the deformation-dependent microstructure of the HPT-disks, position-sensitive high-resolution XRD was carried out on the center, half-radius and edge regions of the specimens processed for N = 3 rotations. The CMWP analysis has revealed that the disks are abundant in lattice defects (~2·10 15 m − 2 ), however, the average dislocation density values do not vary significantly throughout the disks, which clearly indicates that the dynamic equilibrium between dislocation generation and annihilation is already present even after N = 3 rotations. Although the average crystallite size values are practically constant along the radii, a sharper crystallite size distribution corresponding to a more homogeneous nanostructure characterizes the edge region of the MgTN_3h_N3 disk.
It was found that the processing route considerably influences the H-sorption, as the hydrogenation performance of the composite produced for longer co-milling of the titanate nanotubes together with Mg (MgTN_3h_N3) significantly exceeds that of MgTN_0.5h_N3. This observation can be attributed to the different morphology of the additives (better dispersed and partially damaged TN sections) and less to the microstructural parameters of the nanocrystalline Mg phase.