Effect of Mg–Nb oxides on hydrogen sorption kinetics of ball-milled MgH₂

Abstract

Addition of Mg–Nb oxides (e.g. MgNb₂O₆, Mg₄Nb₂O₉, and Mg₃Nb₆O₁₁) ameliorates H₂ absorption/desorption kinetics of MgH₂ as demonstrated in the current article. H₂ desorption and absorption rates of the ball-milled MgH₂ are evidently temperature-dependent, which points out that the prior rate increases with increasing temperature (593–673 K) and vice versa. Among the tested samples, MgH₂ with Mg₃Nb₆O₁₁ nanoparticles showed superior performance. The Johnson–Mehl–Avrami equation was employed to construct H₂ desorption curves as well as find out reaction rate constants at different temperatures. The Arrhenius equation was fitted in the context to estimate the activation energy of the ball-milled MgH₂ and MgH₂/Mg₃Nb₆O₁₁ mixtures; for example, the values obtained were 127 and 88 kJ·mol⁻¹, respectively. In addition, a novel experimental setup combining a hydrogen detector with a differential scanning calorimeter was used to confirm the H₂ desorption properties of the ball-milled nanoparticles discussed based on the kinetic argument.

Graphical Abstract

Open in new tabDownload slide

1. Introduction

The sluggish reaction kinetics of hydrogen (H₂) absorption and desorption in magnesium hydride (MgH₂) under moderate conditions remains a significant obstacle to achieving “zero-emission” hydrogen-fueled vehicles—one of the critical challenges for the future of hydrogen-powered transportation [1–5]. To address this, ball-milling has been identified as a key method for enhancing adsorption/desorption kinetics of absorbent materials. Huot et al. [6] studied the structural differences between ball-milled and unmilled MgH₂. In addition, it has been suggested that transition metals and their oxides serve as crucial catalysts in the interaction of H₂ with Mg. Experimental results indicate that transition metal oxides exhibit higher catalytic activity compared to their corresponding metals [7]. Moreover, coupling oxides such as Cr₂O₃ and Nb₂O₅ has shown greater effects on the process than individual oxides [8]. Barkhordarian et al. [7] found that Nb₂O₅, among transition metal oxides, demonstrated the most promising catalytic impact on the sorption properties of MgH₂. Milling MgH₂ with Nb₂O₅ enhances the adsorption/desorption kinetics through the formation of ternary Mg–Nb oxides after repeated H₂ sorption cycles [9–11]. A proposed mechanism [12] suggests that Nb₂O₅ is reduced to metallic Nb, leading to the creation of Mg–Nb oxides with varying stoichiometry.

Recently, mechanochemically synthesized Mg-based composite materials have shown promising characteristics for re-hydrogenation at room temperature. While intermetallic compounds are known to require challenging conditions for mechanochemical synthesis, their role as nanocatalysts and additives has been well documented in the literature. The exact dissociation mechanism of H₂ molecules on solid surfaces remains an open question, but catalysts/promoters such as TiH₂, TiO₂, Nb₂O₅, MgNb₂O₆, Mg₄Nb₂O₉, Mg₃Nb₆O₁₁, and Ti₄Fe₂O_x have been found to lower the activation energy of hydrogen sorption. These catalysts also facilitate the recombination of atomic hydrogen into H₂ molecules during desorption from MgH₂. In the hydrogenation process of Mg, catalysts help generate atomic hydrogen, while during desorption from MgH₂, they enable the release of H₂ at lower temperatures. For instance, the intermetallic alloy Zr₃V₃O_0.6, stabilized by an oxygen compound with a structure similar to Ti₂Ni, exhibits H₂ sorption properties comparable to Ti₄Fe₂O_x [13].

The H₂ uptake and release properties of binary and ternary Mg–Nb–O compounds have been investigated, with Mg₃Nb₆O₁₁ showing remarkable reactivity with H₂ [14, 15]. Recently, our research group [16] demonstrated the effects of Mg–Nb oxides on H₂ absorption kinetics in Mg/MgH₂ through in-situ X-ray diffraction studies. To further understand the catalytic effects of Mg–Nb oxides on H₂ absorption and desorption reactions in MgH₂, specific studies on the interaction of H₂ with the bare additives were conducted. Dolci et al. [17, 18] were the first to examine the H₂ interaction with bare Nb₂O₅ and WO₃; however, the precise catalytic role of transition metal oxides (binary and ternary) in the H₂ absorption and desorption kinetics of MgH₂, as well as the reaction mechanism, remains unresolved.

In this study, H₂ absorption and desorption kinetics of MgH₂ enhanced by ball-milling with 1 mol% MgNb₂O₆, Mg₄Nb₂O₉, and Mg₃Nb₆O₁₁ were investigated through H₂ desorption measurements. The goal is to better understand the H₂ absorption and desorption mechanisms in the MgH₂/Mg–Nb–O systems. The activation energy of the ball-milled mixtures was estimated using the Arrhenius equation for H₂ desorption and confirmed through direct calorimetric measurements.

2. Experiment

As-milled MgH₂, in the presence of various as-prepared additives such as MgNb₂O₆, Mg₄Nb₂O₉, and Mg₃Nb₆O₁₁ compounds [19], was studied through several experiments detailed in this context. The ball-milled mixtures with these ternary oxides were labeled according to the additives: Nb₂O₅ (O₅), MgNb₂O₆ (O₆), Mg₄Nb₂O₉ (O₉), and Mg₃Nb₆O₁₁ (O₁₁). Additionally, commercial MgH₂ (Comm) and ball-milled pure MgH₂ (BM) were also defined as well. All chemicals were reagent grade and purchased from Sigma-Aldrich, USA.

2.1 Calorimetric analysis

The H₂ desorption properties of the ball-milled samples were measured using a Perkin-Elmer DSC7 differential scanning calorimeter (DSC) connected to an H₂ detector. The instrument was calibrated by flowing a 7.94% H₂/Ar mixture and pure Ar flow to the detector, with the gas flow from the DSC to the detector being monitored. A mixture of H₂/Ar and pure Ar flowed into the DSC, and the corresponding signals were associated with exothermic and endothermic peaks. Since the H₂ detector was placed slightly far from the DSC, a delay in recording data required calibration. The time needed for the signals to travel from the DSC to the detector was calculated to be 231 seconds. Two successive runs were performed for the DSC measurements, and the baseline was constructed by subtracting the data from the second run’s signals.

2.2 Kinetics study

Hydrogen absorption/desorption kinetics, as well as pressure–composition isotherms (PCI), were examined using an Advanced Materials Corporation (Pittsburgh, PA, USA) volumetric Sievert apparatus, with approximately 0.6 g of material used for each measurement. The PCI runs were performed in a gas reaction controller at different programmed temperatures under isothermal conditions for about 45 hours, with a hydrogen pressure of 2.5 MPa for absorption and 0.1 MPa for desorption. The H₂ desorption measurements were carried out at 673 and 653 K with deactivated samples, and at 593–633 K with activated samples. The as-received materials, which interacted with H₂, are referred to as deactivated, while successive PCI runs of the as-received samples are defined as activated. The amount of H₂ in the samples was determined by measuring both the pressure and the sample weight.

3. Results and discussion

3.1 PCI characterization

The results of the PCI experiment for the H₂ desorption curves of the Comm, BM, O₅, O₆, O₉, and O₁₁ samples from 673 to 593 K at 0.1 MPa H₂ pressure are reported in Fig. 1a–1f. For each sample, five lines are presented. The characteristics of the curves demonstrate the expected fast desorption for O₁₁, while Comm exhibited the slowest behavior, with the others falling in between. The desorption curves followed the order: @673 K (□) > @653 K (○) > @633 K (Δ) > @613 K (▽) > @593 K (☆). Comm displayed very slow kinetics even at higher temperatures, releasing almost no hydrogen at 593 K after 100 minutes of exposure (Fig. 1a). BM showed faster kinetics than Comm, and the rest of the materials behaved similarly. Notably, O₁₁ took only 3 minutes for complete desorption. The superior desorption performance of the O₁₁ mixture may be attributed to the presence of octahedral Nb clusters in the Mg₃Nb₆O₁₁ structure [13, 14].

The PCI results for the H₂ absorption curves of different samples at the same temperature range (593–673 K) and 2.5 MPa H₂ pressure are shown in Fig. 2a–2f. For each sample, five patterns are presented. The absorption behavior of the samples followed a similar trend, with lower temperatures being more favorable for absorption. The absorption curves followed the order: @613 K (▽) > @593 K (☆) > @633 K (Δ) > @653 K (○) > @673 K (□). However, the effect of temperatures is very competitive at the test temperatures. Comm exhibited poor kinetics (Fig. 2a), which might be due to a thermodynamic explanation discussed in relation to activation energy. The absorption results for the O₅, O₆, O₉, and O₁₁ samples were comparable to each other, although O₁₁ demonstrated faster kinetics than the others.

The H₂ absorption behavior of the MgH₂/Mg–Nb–O systems is influenced by temperature, with lower temperatures favoring H₂ absorption and higher temperatures favoring H₂ desorption. This means that at lower temperatures, the MgH₂/Mg–Nb–O nanostructure more readily absorbs H₂ due to thermodynamic and kinetic conditions that make H₂ more stable within the material. Conversely, at higher temperatures, the system is more inclined to release H₂, as increased thermal energy allows the hydrogen atoms to overcome the binding forces and exit the material more easily. This temperature-dependent behavior is essential for applications that require reversible H₂ storage, as it enables control over absorption and release by adjusting the temperature. The storage capacity of all samples was lower than the theoretical quantities. For example, O₁₁ absorbed about 95% of its maximum capacity (3.17 wt%) within 10 minutes (Fig. 2f).

Figure 3a and 3b clearly present the effect of additives at the best-performing temperatures (673 and 613 K). O₁₁ exhibited better desorption and absorption kinetics compared to the other ball-milled mixtures. Based on the theoretical H₂ storage capacity of MgH₂ (7.6 wt%), the capacities of the ball-milled samples were estimated using the phase composition of MgH₂ in the ball-milled mixtures, obtained via the Rietveld method (Table 1). The theoretical capacities (wt%) of the various mixtures after ball-milling were as follows: Comm (7.06), BM (6.53), O₅ (6.15), O₆ (5.7), O₉ (5.62), and O₁₁ (5.54). The ball-milled samples were oxidized (MgO ~4 wt%), which significantly affected the reaction kinetics. The amount of H₂ absorbed in the MgH₂/Mg–Nb–O systems was lower than the theoretical storage capacity (7.6 wt%) due to the presence of a non-reactive MgO layer on the surface of the solid materials or at the grain boundaries during ball-milling [14, 15].

In fact, MgO exhibits a peculiar role in H₂ storage kinetics, often acting both as a barrier and, in certain cases, as a catalyst. On one hand, MgO can impede H₂ absorption and desorption by forming a passive layer over MgH₂ particles, which reduces available active sites and increases the energy barrier for H₂ diffusion. This layer restricts the rapid exchange of H₂, potentially slowing down sorption kinetics. However, some studies suggest that MgO, when finely dispersed, can enhance kinetics by providing additional nucleation sites for hydrogenation and dehydrogenation. In this way, MgO could potentially facilitate H₂ diffusion along the MgO grain boundaries and accelerate the breaking and forming of H₂ bonds. The variability in these effects likely depends on factors such as the distribution, thickness, and structural properties of MgO in the MgH₂ matrix, leading to differing views on its role in H₂ storage systems.

The reversible and pronounced interaction of H₂ with Mg₃Nb₆O₁₁ has been explained by a unique structural feature of the solid—specifically, the presence of octahedral niobium clusters within the structure (Fig. 4). These clusters occupy distinct structural positions within the Mg₃Nb₆O₁₁ lattice, which are absent in the other two solids (MgNb₂O₆ and Mg₄Nb₂O₉) examined previously [15]. These niobium clusters could serve as insertion sites for H₂ uptake and release, potentially accounting for the distinctive properties observed.

3.2 Estimation of activation energy

To elucidate the mechanism of catalysis by transition metal oxides, the apparent activation energies for the desorption reaction of MgH₂ catalyzed with Mg–Nb oxide additives were estimated for activated samples using kinetic reaction rate constants at three different temperatures (593, 613, and 633 K). Additionally, an experiment with the deactivated sample was conducted at 653 and 673 K. The activation energy (Ea) for desorption was determined using the Arrhenius Equation (1):

Where k and A correspond to a temperature-dependent reaction rate constant and the pre-exponential factor, respectively. R and T are the gas constant and the absolute temperature, respectively.

Before estimating the activation energy, the value of k is determined from the isothermal H₂ desorption curves. The kinetic rate constant is associated with the reaction mechanisms or power law models subjected to driving forces and geometries of the solids. The desorption mechanism can also be inferred from fitting the data. The Johnson–Mehl–Avrami (JMA) Equation (2) is widely used for kinetic rate estimation of Mg-based materials by fitting the desorption curves [6, 20, 21]:

Where α corresponds to the phase proportion that reacted at time t, and n is the Avrami exponent which generally indicates the morphology of nuclei growth and phase nucleation. Using the JMA equation, n and k were determined by fitting the desorption curves, which clearly demonstrate the PCI desorption data at various temperatures at which the experiment was conducted (Fig. 1). The data extracted from the solid-state reaction mechanism model was then plotted according to Arrhenius equation, i.e. lnk as a function of the inverse of T (Fig. 5).

A good linearity between lnk and 1/T was obtained, and the activation energies were estimated from the slope of the linear fits (Table 1). The apparent activation energy calculated was about 161 and 127 kJ·mol⁻¹ for Comm and BM MgH₂, respectively, highlighting the effect of nanostructuring. The Ea value of activated Comm MgH₂ was close to the value published in literature (156 kJ·mol⁻¹) [22]. Furthermore, the activation energy obtained for the activated O₅ sample was about 66 kJ·mol⁻¹ as reported in the past (62 kJ·mol⁻¹) [23]. The dramatic change in Ea between the deactivated and activated properties of O₅ during PCI measurement suggests that after several H₂ absorption/desorption cycles, Nb₂O₅ becomes highly reactive with the MgO present in ball-milled samples, possibly leading to the formation of ternary Mg–Nb oxides [12] Patah et al. [8] obtained an Ea value of 197 kJ·mol⁻¹ for the dehydrogenation of MgH₂ + 1 mol% Nb₂O₅ ball-milled mixture, which is much higher than that of the activated sample. The dramatic change in Ea between deactivated and activated properties of O₅ during PCI measurement suggests that after several H₂ absorption/desorption cycles, Nb₂O₅ becomes highly reactive with the MgO present in ball-milled samples, possibly leading to the formation of ternary Mg–Nb oxides [12].

The activation energies of O₆, O₉, and O₁₁ were calculated for the first time. We reported that O₁₁ is the best mixture in terms of faster H₂ absorption/desorption kinetics [13, 14]. The Ea value of O₁₁ was expected to be lower than those of the other examined additives, and it was found to be about 39 kJ·mol⁻¹ less than that of the bare additive (BM). In addition, the additives significantly reduced the Ea values and demonstrated a great improvement in kinetics. However, it is important to note that activation energy is not directly related to H₂ desorption kinetics; it is a barrier that must be overcome to initiate the release of H₂. The rate of H₂ release also depends on other factors, such as catalytic effects, surface area, particle size, oxide distribution, and the nature of the oxides added.

3.3 DSC/H₂ detector studies

The results obtained from DSC for Comm, BM, O₅, O₆, O₉, and O₁₁ at different heating rates (5, 10, 20, and 40 K/min) are shown in Figs 6a–11a with the corresponding H₂ detector data displayed in Figs 6b–11b. The desorption curves of Comm MgH₂ have been used as a reference (Fig. 6a and 6b). Solids obtained after ball-milling, characterized by X-ray diffraction (XRD) and analyzed using the Rietveld method, show two types of crystal structures: β-MgH₂ and γ-MgH₂. The metastable orthorhombic γ-MgH₂ is a high-pressure polymorphic form of tetragonal β-MgH₂ [24] and is commonly observed after ball-milling commercial MgH₂ [6, 15].

A single endothermic H₂ desorption peak for each heating rate was observed by DSC for unmilled coarse MgH₂ particles, with peak temperatures (T_p) at 735, 759, 773, and 793 K for 5, 10, 20, and 40 K/min, respectively. The single peaks observed for Comm MgH₂ indicated the presence of β-MgH₂, with no γ-phase in the commercial hydride [16]. However, ball-milled samples contained both phases. A common feature of all curves was the presence of double H₂ desorption peaks, known as a doublet, and the H₂ detector signals were relatively broader than those from DSC, likely due to the time effect.

The H₂ desorption maxima and onset temperatures shifted slightly for all investigated systems, within 20 K. The desorption peaks gradually shifted to higher temperatures with increasing heating rates. However, the effects of particle size and catalytic activity were evident from the shift in peak positions to lower temperatures. The onset temperature (T_on) of Comm MgH₂ at 5 K/min was about 680 K, while that of BM MgH₂ was about 640 K due to the reduction in particle size.

In addition, the T_on of all ball-milled mixtures with additives at 5 K/min was found to be 10–20 K less than for BM MgH₂, verifying the effect of the catalysts. The shifting of desorption peaks in our case is consistent with the results (10 K) reported by Borgschulte et al. [25]. Varin et al. [26] reported that at a specific heating rate (4 K/min), desorption peaks at γ-positions significantly shifted (approximately 10 K) compared to β-positions, which agrees with our results. These phenomena are strongly related to desorption kinetics. The Ea for dehydrogenation of Comm, BM, O₅, O₆, O₉, and O₁₁ was estimated using the Kissinger method [23], according to the following Equation (3):

Where β is the heating rate; T_p is the peak temperature in the DSC curves; R is the gas constant; and k₀ is the frequency factor. In this work, T_p was obtained for several heating rates (5, 10, 20, and 40 K/min). Kissinger plots, i.e. ln(β/T²_p), as a function of the inverse of T_p are shown in Fig. 12. The Ea values were estimated from the slope of the linear fit of data obtained from DSC and the H₂ detector. Prior to fitting, the data were averaged to minimize instrumental errors, mainly due to the delay in the gas flow from the DSC to the detector. The desorption data recorded by the H₂ detector at 40 K/min were relatively scattered and were excluded from the Kissinger plots to improve the fit. The Ea values are shown in Table 1. The Ea for O₆ (180 kJ·mol⁻¹), O₉ (170 kJ·mol⁻¹), and O₁₁ (165 kJ·mol⁻¹) are close to each other; however, these samples showed different results for the H₂ desorption kinetics (Fig. 1).

The Ea value obtained for deactivated Comm was about 164 kJ·mol⁻¹ (from DSC) (reported value 153 kJ·mol⁻¹ [21]) and that of the activated sample was about 161 kJ·mol⁻¹ (from PCI), pointing out that commercial MgH₂ was not very active during H₂ absorption/desorption cycles. In contrast, the BM sample is highly reactive, with Ea reduced by about 30 kJ·mol⁻¹. The activation energy of BM MgH₂ (163 kJ·mol⁻¹) is very close to that of Comm MgH₂ (164 kJ·mol⁻¹), likely due to the presence of an MgO layer on the surface of the ball-milled samples, which requires much energy to overcome. A key point in our findings is that the highly reactive niobia interacts with MgO after several H₂ desorption cycles to form ternary Mg–Nb oxides. The comparison of activation energies obtained from the Arrhenius and Kissinger plots showed some differences. However, the results are comparable to published papers. The Ea obtained for activated Comm MgH₂ and Nb₂O₅-catalyzed samples corresponds to about 161 and 66 kJ·mol⁻¹, as reported in the literature (156 and 62 kJ·mol⁻¹, respectively) [27]. The Ea values obtained from DSC for these two cases are about 164 and 132 kJ·mol⁻¹, respectively.

The change in activation energy between deactivated and activated O₅ observed from DSC and PCI showed a large deviation (66 kJ·mol⁻¹), which is favorable for kinetic arguments. For the other samples, the variation in Ea between the two methods was comparable. Moreover, there is no available literature data related to O₆, O₉, and O₁₁. Several factors may explain the deviation in Ea from the two different results. For example, the source of data for the Arrhenius plot was PCI measurements conducted at 593–673 K under isothermal conditions, whereas Kissinger plots were primarily based on calorimetric measurements conducted from 313 to 973 K at different heating rates.

Mg–Nb oxide additives have a positive impact on the kinetics of H₂ uptake and release by reducing the surface resistance of oxide layers and exposing fresh Mg surfaces to H₂. H₂ diffusion is predominantly influenced by crystal defects and grain size, which serve as nucleation sites for hydride formation (Mg + H₂→MgH₂). In fact, Mg₃Nb₆O₁₁ compound is more active for H₂ absorption at low pressure compared to Nb (V)-based compounds [14–17]. The reversible interaction of H₂ with Mg₃Nb₆O₁₁ has tentatively been explained by the presence of octahedral Nb clusters [14, 15] in the solid structure. The catalytic role of binary oxides (e.g. CuO and Al₂O₃) was also discussed in our previous work [19, 28, 29]. The combination of kinetic models and advanced experimental techniques, such as integrating a H₂ detector with DSC, confirmed the enhanced desorption properties and superior catalytic performance of Mg₃Nb₆O₁₁. It is worth noting that H₂ desorption kinetics is related to factors such as particle size, presence of γ-phase in MgH₂ (which is more active than the β-phase), the MgO layer, the unique activity of Nb₂O₅, and the additional effect of catalysts [29, 30].

Nanostructuring plays a crucial role in enhancing the H₂ storage performance of MgH₂, as it directly impacts both H₂ sorption kinetics and activation energy. Reducing the particle size to the nanoscale increases the surface area-to-volume ratio, exposing more active sites for H₂ interactions and facilitating faster H₂ absorption and desorption. Additionally, nanostructuring reduces the diffusion path for hydrogen atoms within the MgH₂ lattice, enabling quicker diffusion of H₂ through the material. Smaller particle sizes also help relieve strain within the crystal structure during H₂ cycling, which reduces energy barriers and, thus, lowers the activation energy required for hydrogen release. Consequently, BM MgH₂ into nanoparticles creates favorable conditions for H₂ storage, improving kinetic performance by accelerating reaction rates and reducing activation energy.

In a nutshell, hydride-based hydrogen storage materials, with their efficient H₂ capture and release capabilities, can complement renewable energy sources such as biodiesel [31], piezoelectric nanogenerators [32], and solar cells [33] by providing reliable and sustainable energy storage solutions. When paired with biodiesel, these materials can support hybrid energy systems, where stored H₂ can be used to produce electricity or fuel, compensating for biodiesel’s limitations in certain applications. In piezoelectric nanogenerators, which generate power through mechanical stress, hydride storage materials can act as energy reserves to supply power when mechanical energy input is insufficient. Likewise, integrating hydrogen storage materials with solar cells enables the capture of excess solar energy during peak sunlight hours, storing it as H₂ for use when sunlight is unavailable, thus enhancing the stability and efficiency of solar-powered systems. Together, these integrations present a comprehensive approach to achieving a balanced, sustainable energy infrastructure.

4. Conclusions

H₂ absorption and desorption kinetics of MgH₂ have been significantly improved by adding 1 mol% MgNb₂O₆, Mg₄Nb₂O₉, and Mg₃Nb₆O₁₁, as examined by a PCI apparatus. The conclusions of this study are as follows:

1. In the absorption/desorption cycles, the desorption rate of H₂ increases with increasing temperature.

2. The Mg₃Nb₆O₁₁-doped MgH₂ nanoparticle showed the most promising kinetic performance, comparable to MgH₂ promoted with Nb₂O₅, which is reported as the best additive in the literature.

3. In the hydrogen absorption/desorption cycles, the MgH₂/Mg₃Nb₆O₁₁ system completely dehydrogenated (3.93 wt%) within 5 min (at 673 K under 0.1 MPa H₂) and fully re-hydrogenated (3.20 wt%) (at 613 K under 2.5 MPa H₂) in only 3 min.

4. The estimated activation energy for hydrogen desorption for the MgH₂/Mg₃Nb₆O₁₁ mixture and ball-milled MgH₂ were 88 and 127 kJ·mol⁻¹, respectively.

5. A kinetic model demonstrated the role of ternary Mg–Nb oxides on the H₂ absorption/desorption properties of MgH₂ nanopowder.

Acknowledgement

The authors acknowledge the technical support of Professor Marcello Baricco of the University of Turin, Italy.

Author contributions

Md. Wasikur Rahman (Conceptualization [equal], Data curation [equal], Formal analysis [equal], Investigation [equal], Methodology [equal], Writing—original draft [equal]), Muhammad Nurunnabi Siddiquee (Formal analysis [equal], Software [equal], Validation [equal], Writing—review & editing [equal]), Mohammad Mozahar Hossain (Formal analysis [equal], Validation [equal], Writing—review & editing [equal]), and M. Jasim Uddin (Project administration [equal], Resources [equal], Validation [equal], Writing—review & editing [equal])

Conflict of interest statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

This research did not receive any particular awards from subsidizing organizations in general society, business, or non-revenue-driven areas.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

References