Ball-milled samples
Figure 1 shows the X-ray diffraction (XRD) patterns of the starting materials NaH (Fig. 1a) and NaAlH4 (Fig. 1b) together with as-milled NaAlH4 containing samples. It was observed that ball milling (4 h) of NaAlH4 in the presence of 4 mol% TiCl3 resulted in the partial decomposition of NaAlH4 as diffraction peaks of Na3AlH6, NaCl, and Al were found (Fig. 1c). Synthesis of Na3AlH6 was best achieved through ball milling of 2 mol of NaH and NaAlH4, in the presence of 8 mol% of Al and AC (Fig. 1e) which showed (in contrary to Fig. 1d) no diffraction peaks of remaining NaH. Although TiCl3 has been successfully applied as a dopant for sodium alanate systems, the catalytic mechanism remains unclear. It has been reported that during ball milling the titanium of TiCl3 is reduced to its metallic state [12]. However, addition of metallic titanium to NaAlH4 resulted in lower dehydrogenation capacities, indicating that the reduction in the dopant plays more complex role than presumed. The same applies for the Al–Ti compounds that have been proposed as the active species (rather than Ti), but when directly added, the similar results as for non-doped samples were obtained. Bogdanović et al. and Singh et al. reported that although the formation of NaCl as a by-product reduces the overall reversible hydrogen capacity, it acts as a grain refiner for NaH. Similarly, AlxTiy species have been found to act as a grain refiner for Al as well, preventing particle growth [12,13,14]. The presence of grain refiners in combination with the vacancies and defects created during the ball-milling process seems to be crucial. It improves the mass transfer during hydrogenation and dehydrogenation, allowing faster kinetics. The absence of grain refiners would explain the poor results obtained for systems in which metallic Ti or Al–Ti was directly added [14].
In order to investigate the doping effect of the various additives, pre-synthesized Na3AlH6 was used (Fig. 2). XRD measurements showed that addition of 4 mol% TiCl3 alone did not result in a significant decomposition of Na3AlH6 (Fig. 2b), the weak NaAlH4 diffraction peak at 2θ values of 22° was already present in the pre-synthesized Na3AlH6 (Fig. 2a). Furthermore, addition of 8 mol% of both Al and AC showed no appearance of new phases (Fig. 2c). In fact, samples (b) and (c) are very similar.
Sandrock et al. [15] reported that although TiCl3 enhances the dehydriding and hydriding kinetics for both NaAlH4 and Na3AlH6, the effect is more pronounced for NaAlH4 than for Na3AlH6 (activation energy (desorption) for non-catalyzed NaAlH4 = 118 kJ mol−1 (H2), catalyzed (4 mol% TiCl3) NaAlH4 = 80 kJ mol−1 (H2), non-catalyzed Na3AlH6 = 121 kJ mol−1 (H2), catalyzed (4 mol% TiCl3) Na3AlH6 = 98 kJ mol−1 (H2)). From the obtained XRD data, it seems that Na3AlH6 is less reactive toward the TiCl3 catalyst than NaAlH4, as no partial decomposition of Na3AlH6 was observed (i.e., formation of NaCl and NaH). This was further emphasized by in situ pressure recording measurements during ball milling (Fig. 3) which showed that the pressure (from released hydrogen during the reduction of TiCl3) inside the milling vial increased to 5.2 bar for catalyzed NaAlH4, but only to 0.7 bar for catalyzed Na3AlH6. These results allowed for rough estimation of the molar quantity of NaCl produced assuming all TiCl3 has reacted, giving overall Eq. (3) and reversible hydrogen storage of 2.46 wt% for NaAlH4 + 2 NaH + 4 mol% TiCl3 as shown in Eq. (4). For the additive containing sample, the Al addition results in a slightly higher reversible hydrogen capacity (2.61 wt%), see Eqs. (5) and (6).
$$ \begin{aligned} & {\text{NaAlH}}_{4} + \, 2\;{\text{NaH }} + \, 0.04 \,{\text{TiCl}}_{3} \to 0.88\,{\text{Na}}_{3} {\text{AlH}}_{6} \\ & \quad + \,0.24\,{\text{NaH }} + \, 0.12\,{\text{NaCl }} + \, 0.04\,{\text{TiAl}}_{3} + \, 0.24\,{\text{H}}_{2} \\ \end{aligned} $$
(3)
$$ 0.88\,{\text{Na}}_{3} {\text{AlH}}_{6} \rightleftharpoons 2.64\,{\text{NaH }} + \, 0.88\,{\text{Al }} + \, 1.32\,{\text{H}}_{2} $$
(4)
$$ \begin{aligned} & {\text{NaAlH}}_{4} + \, 2\;{\text{NaH }} + \, 0.04\,{\text{TiCl}}_{3} + \, 0.08\,{\text{Al }} \to \, 0.96\,{\text{Na}}_{3} {\text{AlH}}_{6} \\ & \quad + 0.12\,{\text{NaCl }} + \, 0.04\,{\text{TiAl}}_{3} + \, 0.12\,{\text{H}}_{2} \\ \end{aligned} $$
(5)
$$ 0.96\,{\text{Na}}_{3} {\text{AlH}}_{6} \rightleftarrows 2.88\,{\text{NaH }} + \, 0.96\,{\text{Al }} + \, 1.44\,{\text{H}}_{2} . $$
(6)
Thermal dehydrogenation
Thermal decomposition behavior of the NaAlH4-based samples was investigated using DSC measurements. Figure 4 shows the DSC curves of as-received starting materials NaH (Fig. 4a) and NaAlH4 (Fig. 4b) together with milled samples when heated from room temperature to 450 °C under 1 bar of flowing argon at a heating rate of 10 °C min−1. NaAlH4 in the presence of the TiCl3 catalyst precursor (Fig. 4c) showed small and weak endothermic events at 139 and 396 °C, corresponding to the decomposition of NaAlH4 and NaH which both occurred at much lower temperatures than their respective as-received materials, demonstrating the kinetic effect of the catalyst precursor. The weak and broad endothermic event between 169 and 227 °C is contributed to the decomposition of Na3AlH6 (supported by TGA data, Fig. 7), whereas the cause for the exothermic event at 243 °C remains unclear. The most pronounced effect was observed for the sample containing ball-milled NaAlH4 in the presence of 2 mol of NaH and 4 mol% TiCl3 (Fig. 4d), which showed a double endothermic peak for the decomposition of Na3AlH6 between 225 and 250 °C, followed by the endothermic peak for the decomposition of NaH around 390 °C. No endothermic events corresponding to NaAlH4 were observed, further supporting the full conversion of NaAlH4 into Na3AlH6 during the ball-milling process as observed from XRD data (Fig. 1d). Further addition of Al and AC (Fig. 4e) resulted in a similar curve as obtained for the TiCl3-doped sample (Fig. 4c).
Figure 5 shows the DSC measurements for the Na3AlH6 containing milled samples which displayed no significant change for the thermal events, as found for the NaAlH4 containing samples, when heated from room temperature to 450 °C under 1 bar of flowing argon at a heating rate of 10 °C min−1. As-received Na3AlH6 showed two main endothermic events, the first is a complex event between 243 and 304 °C, corresponding to the decomposition of Na3AlH6 into elemental Al and NaH. The second endothermic signal which occurred at 385 °C was caused by the decomposition of NaH. For the TiCl3-catalyzed Na3AlH6 sample (Fig. 5b), the endothermic signal for the Na3AlH6 decomposition step was observed at much lower temperatures (185–278 °C), whereas the endothermic peak for the decomposition of NaH took place at a higher temperature (391 °C). However, the lowest peak temperatures were obtained for the catalyzed Na3AlH6 sample with additional 8 mol% of Al and AC (Fig. 5c) with the Na3AlH6 decomposition reaction between 177 and 265 °C and the NaH decomposition peak at 370 °C.
Na3AlH6 is known to undergo phase transformation of α-Na3AlH6 to β-Na3AlH6 around 252 °C [15, 16], and in order to detect this phase transformation rapid heating and fast data acquisition is needed [17]. Weidenthaler et al. [18] investigated the phase transformation using in situ DSC and high-temperature X-ray diffraction methods and found that using different heating rate resulted in different decomposition behavior. For example, they observed that the endothermic signals for the decomposition of Na3AlH6 and NaH shifted to higher temperatures with higher heating rates. A shoulder (~ 248 °C) at the Na3AlH6 decomposition signal became apparent at heating rates > 10 K min−1, suggesting that another reaction is taken place prior to the decomposition of Na3AlH6 [18]. Figure 5a shows a shoulder at 256 °C which is believed to be the phase transformation of α-Na3AlH6 to β-Na3AlH6. However, in the doped samples (Fig. 5b, c) this shoulder is not observed and is most likely due to increased kinetics resulting in the decomposition of α-Na3AlH6 to NaH and Al without detectable formation of β-Na3AlH6.
The dehydrogenation curves for the NaAlH4- and Na3AlH6-based samples are shown in Figs. 6 and 7, respectively, with their corresponding hydrogen release capacities listed in Table 1. As-received NaAlH4 (Fig. 6a) showed an onset dehydrogenation temperature of 189 °C and hydrogen release capacity of 5.5 wt%, which is close to its theoretical capacity of 5.6 wt%. However, addition of 4 mol% of TiCl3 (Fig. 6b) resulted in an onset dehydrogenation temperature of 90 °C, a shift of roughly 100 °C compared to the pristine material, releasing hydrogen capacity of 4.2 wt%. For Na3AlH6, the effect of TiCl3 was less pronounced (Fig. 7b) and resulted in a shift of roughly 50 °C to lower temperatures and hydrogen release capacity of 2.3 wt% compared to 2.4 wt% for the pristine Na3AlH6 (Table 1). From XRD data (Fig. 2b), this small effect was expected as no change was observed compared to pristine Na3AlH6.
Table 1 Theoretical and experimental hydrogen capacities for the NaAlH4- and Na3AlH6-based samples
For the sample containing both 4 mol% TiCl3 and 2 mol NaH (Fig. 6c), the onset dehydrogenation temperature shifted to higher temperatures in comparison with the sample containing 4 mol% TiCl3 (Fig. 6b) and the sample containing 4 mol% TiCl3, 2 mol NaH, 8 mol% Al, and 8 mol% AC (Fig. 6d) and occurred at 206 °C, providing a hydrogen release capacity of 2.0 wt% (Table 1). From XRD data (Fig. 1d), it could be observed that Na3AlH6 which was synthesized in situ through the addition of NaH resulted in a lower onset dehydrogenation temperature than found for as-received Na3AlH6 (233 °C), Fig. 7a. In Fig. 6d, it can be seen that the hydrogen release capacity increased to 2.5 wt% (Table 1) when 8 mol% of Al and AC were added to the NaAlH4 + 2 NaH + 4 mol% TiCl3 sample and gave an onset dehydrogenation temperature of 149 °C. Similar results were obtained for the doped Na3AlH6 sample when 8 mol% of both Al and AC were added, showing an onset dehydrogenation temperature of 171 °C and a hydrogen release capacity of 2.6 wt%. It seems that, for both the NaAlH4- and Na3AlH6-based samples, the Al and AC additives reduced the onset dehydrogenation temperatures further, possibly by increasing the dehydrogenation kinetics and the increased hydrogen release capacity through addition of Al as proposed in Eqs. (5) and (6).
Cycle stability measurements
Figure 8 shows the cycle stability measurement for the Na3AlH6 + 4 mol% TiCl3 sample when subjected to 2-h hydrogenation (28 bar H2)/dehydrogenation (0.1 bar H2) cycles at 170 °C. The sample showed good cycle stability with a hydrogen capacity of 1.6 wt% at the start of the measurement which increased slightly to 1.7 wt% after 750 h [19]. No capacity loss was observed till 650 h of operation. However, at 650 h the capacity loss of ~ 10–15% which lasted only for a few cycles (Fig. 8) was observed. This capacity loss was not observed for the Na3AlH6 + 4 mol% TiCl3 + 8 mol% Al + 8 mol% AC sample and it’s unclear as to the reason for its occurrence.
Addition of 8 mol% Al and 8 mol% AC to the Na3AlH6 with 4 mol% TiCl3 (same cycling conditions as for the Na3AlH6 + 4 mol% TiCl3 sample) resulted in an increased hydrogen capacity of 2.2 wt% (Fig. 9) in comparison with 1.7 wt% (Fig. 8). The sample showed good cycle stability and in contrary to the Na3AlH6 + 4 mol% TiCl3 sample, no capacity losses were observed during 750 h of operating. Electricity shutdown caused some interruption at around 50, 170, 250, and 480 h since beginning of the experiment.