Tailoring LPSO phases in Mg–Y–Zn alloys to govern hydrogenation kinetics

A series of Mg–Y–Zn alloys with varying long-period stacking ordered (LPSO) phase fractions were prepared through control of alloy content, heat-treatment, and single-pass extrusion. The effect of LPSO phase volume fraction and microstructure refinement on the hydrogen absorption/desorption properties of ball-milled powders prepared from the extruded alloys was experimentally assessed. The hydrogen absorption and desorption kinetics scaled with the LPSO phase volume fraction, though the results of this study suggest that the scaling is not linear. Variations in the LPSO phase fraction and alloy content did not alter the (de)hydrogenation equilibrium pressure, indicating there is no significant change in thermodynamics of hydrogenation. Hydrogen absorption experiments on thin foils made from the extruded Mg–Y–Zn alloy with a high LPSO phase fraction demonstrated that the LPSO structures decompose into Mg phase, lamellar Mg/Mg–Zn structures and YH2 particles at hydrogen partial pressures sufficient to form YH2. This study shows that the hydrogen absorption/desorption kinetics in the Mg–Y–Zn alloys can be controlled by tailoring the LPSO phases using conventional metallurgical techniques.

Over the last twenty years an area of particular research interest has been magnesium alloys that contain both rare-earth elements (RE) and transitions metals (TM), as such alloys have been found to contain unique LPSO structures where the solute elements segregate to form ordered layers at the repeated stacking faults [25][26][27][28][29][30][31][32]. Magnesium alloys with LPSO structures have shown improvement in mechanical properties, including strength, ductility and high temperature creep resistance, depending on the LPSO polytype variants and their volume fractions [33][34][35][36][37]. More recently, studies have examined the potential of LPSO structures to improve the hydrogen properties of magnesium [20,[38][39][40][41][42][43][44][45]. LPSO structures intrinsically combine the catalyst and microstructure refinement methods commonly used to improve hydrogen storage properties. The J Mater Sci (2023) 58:8572-8596 formation of LPSO structures requires elements that are known hydrogen catalysts which are then distributed uniformly in the alloy at the nanoscale during solidification.
Wrought alloys with high fractions of LPSO structures have also been shown to have a high degree of recrystallization with ultrafine grains [33][34][35][36][46][47][48]. Such grains have been observed along the LPSO phase-Mg matrix interface [47,48], at LPSO fragments in the Mg matrix [33] and at grain boundaries and kinks in the LPSO phase [46,48]. Recrystallization volumes increase as the LPSO phase volume fraction increases due to a combination of increased Mg-LPSO interface area for stress concentration, increased LPSO fragment particle density for particle stimulated nucleation (PSN) and increasing pinning of dynamically recrystallized grain boundaries [33][34][35]. Grain refinement of wrought Mg-RE-TM alloys due to deformation of LPSO phase, therefore, has the potential to mass-produce effective hydrogen storage materials using industrial-scale metallurgical processes.
Even with the heightened interest, the discovery of LPSO structures in Mg-RE-TM alloys is relatively recent and much is still unknown or uncertain about their fundamental properties. Different LPSO polytypes, including 24R, 14H, 18R, and 10H, have been reported with their formation influenced by the choice of alloying elements and thermo-mechanical process history [27,30,32,36,37,[49][50][51][52]. Polytypes can coexist directly from solidification or come to coexist through phase transformations during heat treatment and/or precipitation of new phases. The variety of alloy compositions and processes that can be used to vary LPSO structures and phase volume fractions means there is a considerable range of alloys and conditions that have yet to be systematically assessed.
Even less is known about the hydrogen properties of magnesium alloys with LPSO structures. Most of the available studies utilize only a single alloy composition and/or process condition [38-41, 44, 45, 53]. In hydrogenated LPSO alloys, the RE hydrides have been shown to act as catalysts that transport hydrogen into the Mg matrix via a low-energy chain of interstitial sites and vacancies at the hydride-Mg interface, which is known as the hydrogen spill-over effect or hydrogen pump effect [41,54]. Transition metals found to form LPSO structures, including Co, Fe, Ni, Ti, and V, catalyse the dissociation of H 2 [43]. However, the use of solute elements has two competing and interrelated effects on hydrogen properties. Additives can be used to improve kinetics and potentially create lower energy pathways, but at the expense of the gravimetric capacity, as elements denser than Mg increase the average density of the material. It is, therefore, important to study the solute content in LPSO alloys to optimize both hydrogen kinetics and capacity through the effective internal distribution of minimal additives. As the LPSO phase volume fraction is linked to both strategies of ultrafine grain refinement and favourable catalyst additions, a study to correlate LPSO phase volume fractions with hydrogen storage properties in Mg-RE-TM alloys is required.
In the current work, the LPSO phase volume fraction in Mg-Y-Zn alloys has been tailored through chemical composition and thermo-mechanical treatment. Furthermore, it is shown that the LPSO phase volume fraction correlates with the hydrogen kinetics.

Materials preparation
Three ternary compositions (Mg 97 Y 2 Zn 1 (at%.), Mg 92 Y 5 Zn 3 (at%.), and Mg 83.3 Y 9.5 Zn 7.2 (at%.)) were selected, which closely follow the Zn:Y 3:4 composition line seen in Fig. 1 in order to supress the formation of phases other than Mg and the LPSO phase. Alloy castings were performed in an induction furnace at 760°C using 99.9% pure Mg and Zn and a Mg-25 wt.% Y master alloy. The melt was stirred under an argon atmosphere before being poured into a water-cooled cylindrical mould. The as-received alloys were each cut into three sections; one section was retained for characterization in the as-cast condition and are hereafter designated as ZC alloys. 'Z' denotes that zinc is the transition metal used throughout the alloys in this study while 'C' denotes the as-cast condition. The remaining two cylindrical sections of each alloy were heat-treated together.
The Mg-Y-Zn alloy sections were heat-treated under an argon atmosphere in a Labec muffle furnace. Sections were heated from 25°C up to 520°C at a rate of 5°C/min, held at 520°C for 10 h and then directly water-quenched to 70°C. One section of each alloy was retained for characterization in the heattreated conditions designated as ZHT with 'HT' denoting heat-treated alloys. The final section of each alloy was used for extrusion.
Cylindrical billets 29.5 mm in diameter and 20.0 mm in height were machined from the final sections and extruded with a HafCo Metal Master HP-150 industrial hydraulic press equipped with a custom heated sleeve to maintain the temperature in the billet and die. Extrusion was performed at 450°C with a ram velocity of 0.5 mm/s; the billet diameter decreased from 29.5 to 10 mm, which introduced an equivalent strain of 2.16. The extruded samples were dropped into a water bath positioned directly below the sleeve. The extruded samples further are designated as ZE.
To perform hydrogen absorption, desorption, and pressure-composition (PC) measurements, powders of the extruded alloys were prepared by manually rasping sections of the extruded samples into millimetre-sized chips that were then processed by ballmilling. The chips were pulverized in a planetary ball mill for 4 h at 800 rpm with a ball-to-media ratio of 20:1 under a hexane atmosphere to protect the powder from oxidation during milling.
Thin foils of the extruded Mg 83.3 Y 9.5 Zn 7.2 alloy were prepared to study the microstructure after hydrogen-induced decomposition of the LPSO phases while excluding the effects of ball-milling. The foils were thinned to less than 100 lm to hydrogenate as much of the foils as possible, as hydrogen diffuses approximately 30 lm into the surface of magnesium before the forming hydride layer inhibits further uptake [4]. The polished foils and hydrogenated foils were kept in membrane boxes to protect against oxidation.

Characterization techniques
Samples of the as-cast, heat-treated, and extruded alloys were prepared for SEM observation and EDS measurement by cutting plates, including two plates that were parallel and perpendicular to the extrusion direction, and then polished by standard procedures using a Struers Labopol-21 polishing wheel. The thin foils after hydrogenation were also polished to remove the surface layer. Micrographs of the polished alloys and hydrogenated foils were taken with a JEOL JSM-7800F SEM with a spatial resolution of 3.0 nm when operated at 15 kV accelerating voltage, 5 nA probe current, and 10 mm working distance. In this study, a 20 kV accelerating voltage was used for  (designated by  open square, round, triangle,  rhombic symbols and letters  ZC. ZHT, ZE), compared with reported compositions of LPSO polytypes [29,30,61] (designated by star symbols). secondary electron (SE) and backscatter electron (BSE) imaging at working distances of 13.1 mm and 6.0 mm, respectively.
For phase volume fraction measurements, a variant of the manual point count procedure was used. Three SEM BSE micrographs were taken of each alloy in all three process states. The images were processed using the Fiji/ImageJ image analysis application [55]. In this study, grayscale pixels were used as points in the test grid with the total point count equal to all the pixels within the field of view. Grayscale value ranges were assigned to individual phases in SEM micrographs using region of interest (ROI) gray-scale histograms. The volume fraction of each LPSO phase and Mg phase was calculated by correlating the grayscale range with the EDS composition measured from phases within the ROI. Multiple EDS measurements were taken within the area of the SEM micrographs to quantify the content range of a given phase with statistically valid quantitative data [56]. The calculated volume fractions for each phase from three different SEM micrographs were averaged to provide the phase volume fractions in each alloy.
Elemental compositions of phases in the alloys were determined from EDS point scans and line scans using an Oxford Instruments X-max 50 mm 2 EDS detector with 3.0 nm spatial resolution integrated with the JEOL JSM 7800F SEM. In this study, EDS spectra were collected with a 5 kV accelerating voltage to minimize the interaction volume of Mg. The error values for the EDS measurements are too small to be shown in Fig. 1, but the values have been estimated and can be found in [56]. The detector and spectra capture were controlled with Oxford Instruments' AZtec application. Parameters for data capture were tuned so that each spectrum was generated with 1 million detector counts for statistically valid results. The at.% composition data from the EDS spectra were processed using Oxford Instruments' AZtec application and then exported to calculate the average phase compositions.
Foils to study the LPSO phases with transmission electron microscopy (TEM) were prepared from the heat-treated and extruded alloys with typical methods. 3 mm discs were punched from mechanically ground foils of the alloys and then electrolytically thinned with a Struers Tenupol 5 twin jet electropolishing unit operated at 100 V. Selected discs were finished with a Gatan 691 precision ion polishing system (PIPS) using 5 keV and a 4°angle until foil perforation followed by additional polishing for 30 min at 3 keV and a 2°angle. TEM brightfield images of the microstructure and diffraction patterns of the LPSO phases were taken with a FEI Tecnai G 2 T20 Twin TEM equipped with a LaB 6 emitter operated at 200 kV accelerating voltage and an integrated Gatan Orius SC200D CCD digital camera. Additional images were taken using a JEOL JEM 2100 TEM equipped with a LaB 6 emitter operated at 200 kV accelerating voltage and an integrated Gatan Orius SC1000 CCD digital camera.
The samples for X-ray diffraction (XRD) studies were prepared from 3 mm-thick plates of the extruded alloys cut in the same manner as the SEM samples previously described. XRD measurements from the extruded samples were performed parallel to the extrusion direction. XRD patterns were collected with a Malvern Panalytical X'pert3 Pro Powder x-ray diffractometer using a Cu Ka x-ray source over the 2h range between 6.681°and 80.000°with a step size of 0.0131°. The time per step for the alloy plates and hydrogenated foils was 3.28 and 6.56 s, respectively. Phases in the XRD patterns were identified using Malvern Panalytical Highscore Plus software [57] interfaced with the International Centre for Diffraction Data (ICDD) Powder Diffraction File (PDF4 ?) database. For the LPSO phases, d-spacings and the equivalent 2h positions of XRD peaks were simulated with the Highscore Plus software Bragg calculator using space groups and lattice parameters reported by Egusa and Abe [29] for the 18R and 14H LPSO polytypes and Yamasaki et al. [30] for the 10H LPSO polytype.

Hydrogen experiments
Hydrogen absorption and desorption experiments to measure kinetics and thermodynamics were done using a custom-built Sieverts' apparatus [15,[58][59][60]. The apparatus is a closed volume system with a 21 ± 1 cm 3 stainless steel reactor vessel and heated by an external resistance furnace with temperature control of ± 1°C. Hydrogen of 99.99999 at% purity was delivered to the system by desorption from LaNi 4.15 Fe 0.85 H x hydride. Data acquisition was controlled by the LABView application.
Prior to formal measurements, approximately 2 g of each powder was heated in the reactor at 400°C for two hours and then activated by ten rapid (de)hydrogenation cycles using 420°C and 30 atm during adsorption and 280°C and 2 atm during desorption. The hydrogen kinetics of the activated samples were recorded over two (de)hydrogenation cycles at 300°C. Desorption was carried out against vacuum in the reference cell. After the second desorption cycle, pressure-composition measurements were taken at 300°C.
For the thin foil experiments, two foils of the extruded Mg 83.3 Y 9.5 Zn 7.2 (ZE3) alloy with a high LPSO phase fraction were hydrogenated with different hydrogen pressures at 400°C for 20 h. One foil was partially hydrogenated to form only YH x hydrates, which was defined later as YH 2 by XRD analysis, to determine if formation of Y hydrates is sufficient to cause LPSO phase decomposition; this foil is referred to as the partially hydrogenated ZE3-P foil. The second foil was fully hydrogenated to form both YH 2 and MgH 2 and is referred to as the fully hydrogenated ZE3-F foil.

Results and discussion
Tailoring LPSO polytype and volume fraction by composition and processing in Mg-Y-Zn alloys LPSO phase composition Figure 1 plots the ternary composition of the LPSO phases measured by EDS and compares them to the stoichiometric compositions of LPSO polytypes from the literature [29,30,56,61]. The error values for the EDS measurements are too small to be shown in Fig. 1, but the values have been estimated and can be found in [56]. The measured content of the LPSO phases lies between the ideal stoichiometry for the 14H and 18R polytypes with an excess of zinc and deficiency of yttrium, so the phases tend to approach the Zn:Y = 1:1 ratio. While the 3:4 ratio represents the ideal stoichiometry of the LPSO polytypes, LPSO phases have been shown to form with a non-stoichiometric content between 1:2 and 1:1 due to imperfect ordering of the Zn 6 Y 8 clusters and occupation of the cluster interstitial by Mg, so the Zn:Y ratios in this study are not atypical for LPSO structures. The LPSO phases have been assigned polytypes based on proximity to the polytypes on the 1:1 composition line. The composition of the LPSO phase being closest to 14H in the ZC1 and ZC2 alloys is unexpected, as the 18R type has been frequently reported in low content as-cast alloys and CALPHAD calculations have determined that the 14H polytype can only form from solid-state reactions [61].
The Y and Zn content in the LPSO phases increases after heat treatment and extrusion. The Zn content in the LPSO phase increases the most, which indicates Zn has a high affinity for the LPSO phase. The greatest increase in Y and Zn content occurs between the as-cast and heat-treated conditions, with a further increase of lower magnitude after extrusion of the heat-treated samples. The greater increase seen in the heat-treated condition is due to the long diffusion distances of Y and Zn at the temperatures and annealing times employed in the present study. The LPSO phase in the ZE2 alloy demonstrates a noticeable increase in Y and Zn going from the as-cast state to heat-treated and then extruded states. In contrast, heat treatment and extrusion have no significant effect on composition of the 10H phase and slightly increase the Zn content of the 18R phase in the Mg 83.3 Y 9.5 Zn 7.2 alloy. The overall composition change in the LPSO phases in the alloy with high levels of alloying elements is less than the changes in the alloys with lower levels of alloying elements.
EDS measurements of the composition of the Mg phase after each processing step were also performed. The Mg-based matrix in all the Mg 97 Y 2 Zn 1 and Mg 92 Y 5 Zn 3 alloys contained a small amount of yttrium and trace amounts of zinc. The yttrium content in the Mg matrix increases from 1 at% Y in the as-cast Mg 97 Y 2 Zn 1 alloys to 2 at% Y in the as-cast Mg 92 Y 5 Zn 3 alloys, but it remains within the solubility limit of 2.95 at% Y at 400°C. Heat treatment only marginally increases the Y and Zn content of the Mg matrix, while extrusion does not result in any further significant change. The Mg phase was not identified in the Mg 83.3 Y 9.5 Zn 7.2 alloys. The lack of significant composition changes in the Mg 83.3 Y 9.5 Zn 7.2 alloys suggests that the presence of the Mg phase is important for LPSO phase transformation as a source, or sink, for diffusing Mg atoms. This point is discussed further in Sect. ''Phase volume fractions'' after the phase volume fraction results are presented and correlated with the LPSO phase composition changes. Figure 2 compares the microstructure of the three Mg-Y-Zn alloys across the as-cast, heat-treated, and J Mater Sci (2023) 58:8572-8596 extruded conditions. The Mg phase in the ZC1 and ZC2 alloys forms dendritic structures with a similar characteristic size of approximately 10 lm, which grow and coarsen with heat treatment to approximately 50 lm in the ZHT1 alloy and between 10-50 lm in the ZHT2 alloy. Extrusion causes the Mg phase to flatten and elongate along the extrusion direction, leading to a size decrease of the Mg grains in the transverse plane in both the ZE1 and ZE2 alloys. In the longitudinal direction, the Mg grains have elongated in both ZE1 and ZE2 alloys over hundreds of microns while the width of the phase has remained approximately the same as in the corresponding heat-treated alloys. The ZC3 alloy consists almost entirely of LPSO phases and no Mg phase is observed.

Microstructure
The LPSO phases have two distinct morphologies in the as-cast condition. Blocky grains of LPSO phase form a network within the Mg matrix while micron thin LPSO plates are seen within the Mg grains themselves. Figure 2a shows  of microns in size with different orientations. The amount of block structure increases with increasing amounts of alloying elements, while the plates coalesce into lamellar structures that maintain the localized parallel orientation. In the ZC3 alloy, the microstructure consists almost entirely of LPSO phases that extend for hundreds of microns in large monolithic structures. Heat treatment has not changed the size of the LPSO domains, but the spacing between them has increased in the ZHT1 and ZHT2 alloys due to the increase in the volume fraction of the Mg phase. The LPSO laths in the ZHT3 alloy appear more uniform in contrast compared to the ZC3 alloy, but otherwise there is no significant change in microstructure after heat treatment. The backscatter contrast seen in Figs. 2c, f, i indicates that there is a compositional difference within the LPSO phase in the ZC3, ZHT3, and ZE3 alloys, with a higher Y and/ or Zn content in the high-contrast regions. EDS results confirmed that there is a consistent difference in Y and Zn content between the low contrast and high contrast areas. As the contrast and composition difference is seen in the ZC3 alloy, this means that the darker 18R and lighter 10H polytypes both formed in this alloy during solidification. The contrast of the LPSO laths becomes more uniform after heat treatment, indicating that the heat treatment causes composition homogenization of the laths. Moreover, the boundaries between high and low contrast LPSO laths in the ZHT3 alloy are more diffuse after heat treatment, which suggests that the different LPSO types are mixing at the boundaries.
Extrusion has not affected the size of the LPSO domains, but the refinement of the Mg phase reduces the spacing between LPSO domains in the transverse plane. The LPSO domains are elongated for hundreds of microns parallel to the extrusion direction, while their width remains the same as in the heat-treated condition. In the ZE1 and ZE2 alloys, the elongated LPSO domains are fragmented into smaller particles (see Fig. 2j). In the ZE3 alloy the fragmentation is reduced as the amount of block-type LPSO domains increases, with no fracturing observed. Extrusion results in wavy overall morphology of the LPSO structures and produces kinks in some regions, which is most obvious in LPSO structures that contain thin Mg plates. These deformation microstructures are similar to what has been previously observed in LPSO alloys processed by extrusion or ECAP [37,62,63].
The SEM and TEM micrographs of the Mg 97 Y 2 Zn 1 , Mg 92 Y 5 Zn 3 , and Mg 83.3 Y 9.5 Zn 7.2 alloys in the heattreated and extruded conditions are presented in Figs. 3, 4, and 5, respectively. Identical extrusion processing produces different microstructures in the alloys that seem to depend on the pre-existing volumes of Mg and LPSO phases. Figures 3b, c, f show that regions of the ZE1 Mg matrix contain micronsized grains. No refinement of the block and lamellar LPSO structures is observed by SEM after extrusion, but kink deformation has been introduced as seen in the curvature of the Mg/LPSO lamellae in Figs. 3b, e. Kink deformation is also observed in the ZE2 alloy. Grain refinement resulting in micron-sized grains of Mg matrix is seen in the TEM micrographs of ZE2 alloy in Figs. 4e, f, and their size is smaller than that in the ZE1 alloy. The ZE3 alloy shows the most significant microstructure change after extrusion, as recrystallized grains with an average size of approximately 500 nm have formed within the LPSO structures. Figures 5b and 5c show that the fraction of recrystallized grains is high, but there are also monolithic LPSO structures that do not show any recrystallization. These results show that an increasing LPSO volume fraction in Mg-Y-Zn alloys promotes grain refinement during extrusion presumably due to DRX in the Mg matrix and of the LPSO phase itself.
The TEM diffraction patterns of the LPSO phase in the extruded alloys (Figs. 3, 4, and 5) are typical for LPSO phases previously reported in the literature [27,29], where additional reflections are seen between the centre maxima and the Mg (002) reflection. Table 1 summarizes the TEM diffraction results for the LPSO phases in the heat-treated and extruded alloys; the full set of TEM diffraction patterns and the extended analysis for the LPSO phases can be found in [56]. The LPSO polytypes identified in the heattreated and extruded alloys using TEM diffraction nearly all agree with those identified by EDS composition. preparation and data collection being restricted to small random areas of the microstructure.

Phase volume fractions
The most abundant phases in the Mg-Y-Zn alloys are Mg and LPSO, with the LPSO phase increasing at the expense of Mg as the alloy content increases. alloys. The changes are smaller in the Mg 97 Y 2 Zn 1 alloys because the smaller LPSO phase volume fraction would not lead to as much Mg diffusion into the matrix.
In contrast, the Mg 83.3 Y 9.5 Zn 7.2 alloy does not show a similar trend, as the total volume fraction of LPSO phases remains stable at approximately 99%. This stability, when contrasted with the changes in the Mg 97 Y 2 Zn 1 and Mg 92 Y 5 Zn 3 alloys, shows that mobility of Mg atoms is important for changing the phase volume fractions in Mg-Y-Zn alloys. The large change in the Mg 92 Y 5 Zn 3 alloys implies that this is a function of a large contact area between the LPSO phase and Mg matrix, which is a topic that should be further studied.
In addition, there is a small fraction of Y-rich particles, which is below 1% in all the alloys. The particles appear individually and in clusters within the Mg and LPSO phases.

X-ray diffraction
The XRD patterns of the three Mg-Y-Zn alloys in the extruded conditions are compared in Fig. 7. All of the XRD patterns contain peaks from the dominant Mg and LPSO phases; low intensity peaks are also matched to the W-phase (Mg 3 Y 2 Zn 3 ) in the ZE2 and ZE3 alloys. There is also a consistent unmatched peak that is present in all alloys between the Mg peaks with highest intensity. The position of this peak is A limited number of XRD peaks that are distinct from different LPSO polytypes allow for comparison with the EDS results. The 14H phase identified in the ZE1 pattern is consistent with EDS data. The ZE2 pattern contains the 14H and 18R polytypes, which correlates with Fig. 1 where it is seen that the LPSO phase content lies between these two polytypes. The identification of 18R in the ZE1 and ZE2 alloys with XRD also suggests that some amount of 18R LPSO phase was retained in both the ZC1 and ZC2 alloys, but it was not identified by EDS measurements. The ZE3 pattern contains peaks for both 18R and 10H, which is consistent with the two LPSO phases found with EDS. A limited number of peaks are also assigned to 14H, which implies that a small amount of 14H coexists with 18R and 10H in the ZE3 alloy.
The Mg (0002) and (0004) XRD peaks exhibit notably high intensity in all extruded alloys. As the measurements have been taken from surfaces parallel to the extrusion direction, the increased relative intensity of the Mg basal reflections indicates that the extruded alloys all have a basal texture with the basal planes parallel to the extrusion direction. This basal texture is common in extruded Mg alloys, including MgRE alloys with LPSO phases [64]. Additional texture information is difficult to discern, but it is observed that the Mg 2021 À Á peak intensity is relatively low in all of the extruded alloys. In addition, the Mg 101n À Á peaks in the ZE3 pattern are reduced to the point of near-total absence, including the most intense peak for 1011 À Á planes. The low intensity of significant Mg peaks implies there is a texture associated with the LPSO phase. However, the detailed studies of the textures in the Mg-Y-Zn alloys were beyond the scope of the present study.
The mixed results of the alloy characterization show the difficulty in identifying all of the LPSO polytypes that may be present in a non-equilibrium alloy. Figure 1 shows it may be possible for 14H plus 18R or 18R plus 10H to coexist between the Zn:Y ratios of 3:4 and 1:1. The issue of experimental LPSO phase composition is not easily resolved as LPSO  polytypes have been shown to form with a high degree of non-stoichiometric composition due to the accommodation of in-plane lattice disorder [29]. Furthermore, coexisting LPSO polytypes within a LPSO phase region may be difficult to spatially resolve if the different stacking faults are interleaved. As most work with Mg-Y-Zn alloys would not include equilibrium heat treatments, more work is needed to understand what impact a mixture of LPSO polytypes have on alloy properties.

The effect of LPSO phases on hydrogenation properties of Mg-Y-Zn alloys
Absorption/desorption kinetics The hydrogen absorption and desorption kinetics at 300°C for the three extruded Mg-Y-Zn alloys are given in Fig. 8. Measurements were taken after ballmilling and activation of the alloy powders. Desorption measurements in Fig. 8b were taken immediately after the corresponding absorption measurement in Fig. 8a. The combined measurement error associated with the uncertainty in the reactor volume and the pressure sensor error is approximately 10%. All three alloys exhibit similar characteristic absorption curves in Fig. 8a, consisting of three stages. An initial short stage on the order of 1 s with almost no absorption is followed by a rapid stage where each alloy completes 68-85% of the reaction after approximately 200 s. Finally, there is a long period of slower absorption that persists for the remaining time. After 10,000 s (2 h 47 min) the effective hydrogen absorption capacity is 5.5 wt.% for ZE1, 5.9 wt.% for ZE2, and 6.0 wt.% for ZE3. Absorption continues for an extended period afterwards until the alloys achieve a maximum absorption of 6.3 wt.% for ZE1 and 6.6 wt.% for both ZE2 and ZE3. ZE3 demonstrates the greatest absorption in the first 10 s, absorbing 2.0 wt.% of hydrogen compared to 0.5 wt.% for ZE1 and 1.8 wt.% for ZE2. Although the ZE3 alloy contains the most LPSO phase with the associated catalytic elements, the absorption kinetics are comparable with the ZE2 alloy processed in the same manner but containing less LPSO phase. The lag seen in the ZE1 alloy is attributed to fewer catalysts being less effectively distributed on the particle surface.
The desorption curves of all three alloys follow a typical sigmoidal behaviour in Fig. 8b, but there is a pronounced difference in desorption kinetics. The overall desorption rate increases in the order ZE1 \ ZE3 \ ZE2. The initial slow desorption stage is significantly longer for each alloy compared to the initial absorption stage. After 10,000 s ZE2 and ZE3 have desorbed 5.5 wt.% and 5.9 wt.% of hydrogen, respectively, but ZE1 has only released 0.8 wt.% by (c) Mg 83.3 Y 9.5 Zn 7.2 (The peak labelled 'No match' was not indexed to any of the identified phases, but its position between high intensity Mg peaks is attributed to additional reflections produced by the LPSO superlattice). this time. ZE2 desorbs a maximum of 5.6 wt.% at 20,000 s and ZE3 desorbs a maximum of 6.5 wt.%, at 25,000 s. ZE1 never demonstrates a desorption plateau during either cycle even for 100,000 s. At this point during the second cycle the experiment is terminated where ZE1 has desorbed 5.5 wt.% of hydrogen. The order of desorption kinetics is primarily determined by how early the alloy begins to desorb hydrogen, with the earliest desorption by ZE2 determining its fast desorption kinetics.
The difference in maximum capacities between absorption and desorption shows that ZE3 is the most effective at reversibly desorbing hydrogen, as it retains only 0.1 wt.% of the hydrogen that it absorbed after the second cycle. In comparison, ZE1 retains 0.7 wt.% after desorption while ZE2 retains the most at 0.9 wt.%. ZE1 shows the greatest difference in the time needed to reach maximum capacity, as desorption time for ZE1 is ten times greater than for absorption. For ZE2 and ZE3, the time for full desorption by each alloy is twice that needed for absorption.
The effective hydrogen absorption capacities and reaction fractions of the alloys after 200 s are plotted against the LPSO volume fraction in each alloy in Fig. 9a. The similarity in absorption between cycles is once again seen here, apart from ZE3 which shows a significant increase in capacity between cycles. Linear regression shows that increasing LPSO volume fraction is correlated with greater effective absorption capacity and a greater fraction of the hydrogenation reaction is completed by the end of the rapid stage.
The major differences in desorption rates and time to initiate rapid desorption make comparing the Mg-Y-Zn alloys at the end of the rapid stage less feasible. Instead, the desorption capacity and reaction fractions after 10,000 s of desorption are plotted against the LPSO phase fraction of each alloy in Fig. 9b. This period is selected as it is the same length as the time each alloy required to reach the reversible absorption capacity and covers the entire dehydrogenation reaction for the fast-desorbing ZE2 alloy. In this context, effective desorption capacity increases with increasing LPSO phase fraction. However, ZE2 has the fastest kinetics as its reaction fraction is greater than ZE3.
There are not many studies on hydrogen absorption in Mg-Y-Zn alloys to compare the results of hydrogen experiments in this study. Recently Zhang et al. [20] found that chips of Mg 98.5 Y 1.0 Zn 0.5 alloy in as-cast, homogenized, and ECAP process conditions all absorbed approximately 5.2 wt.% H after 10,000 s at 280°C, which increased to approximately 6.5 wt.% at 320°C for the as-cast and ECAP alloys. Hydrogen desorption from the Mg 98.5 Y 1.0 Zn 0.5 alloys was negligible at 280°C, but increased to 5.0 wt.% H at 320°C. Another recent study by Zhang et al. [45] on hydrogen absorption by ball-milled Mg 96 Y 2 Zn 2 alloy found that the powder absorbed 5.4 wt.% H and desorbed 4.5 wt% at 300°C. The Mg 97 Y 2 Zn 1 alloy in this study shows comparable absorption kinetics, but the desorption kinetics are significantly slower. In contrast, the Mg 92 Y 5 Zn 3 and Mg 83.3 Y 9.5 Zn 7.2 alloys show improved hydrogen kinetics when compared to these recent studies.
Zhang et al. [20] found that the Mg 98.5 Y 1.0 Zn 0.5 alloy in the as-cast condition performed better than in the homogenized and ECAP conditions and attributed this improvement to the more uniform distribution of the YH 2 nanocatalysts formed in-situ as the LPSO phase decomposed. Zhang et al. [45] also found that the homogeneous distribution of the YH 2 nanocatalysts played a crucial role in improving the desorption kinetics of the Mg 96 Y 2 Zn 2 alloy. The slow desorption kinetics of the Mg 97 Y 2 Zn 1 alloy in this study is then explained by a poor distribution of the YH 2 nanocatalysts within the decomposed LPSO nanocomposite. Similarly, the fast desorption kinetics of the Mg 92 Y 5 Zn 3 and Mg 83.3 Y 9.5 Zn 7.2 alloys are explained by a good distribution of the nanocatalysts. As the LPSO phase volume fraction has been shown to increase with alloy content, the connection to hydrogen kinetics is the increase in Y that is uniformly distributed throughout the alloy by the LPSO phase for effective formation of the YH 2 nanocatalysts.
Compared to other Mg alloys with LPSO phases, the Mg-Y-Zn alloys in the current study have similar hydrogen absorption and desorption capacities but show slower kinetics around 300°C. Mg-Y-Ni alloys in particular show superior hydrogen absorption and desorption kinetics, based on the summary of hydrogen properties of selected alloys given in Table 2. This is due to the additional catalytic effect of nickel.

Pressure-composition isotherms
After the cyclic absorption and desorption experiments are complete, pressure-composition (PC) measurements are taken of the Mg-Y-Zn powders. The hydrogen absorption and desorption isotherms at 300°C are presented in Fig. 10a. All three alloy powders produce a broad and flat pressure plateau with low pressure hysteresis. These features indicate that the MgH 2 formation/dissociation reaction is highly reversible in the extruded Mg-Y-Zn powders at 300°C. No additional plateaus are observed in the pressure curves, which indicates that no other hydride reactions have occurred [67].
The hydrogen plateau pressures determined in the present work are summarized in Table 3. The plateau pressures are reported as a range with an approximate upper and lower bound. Within the experimental error, absorption and desorption plateau pressures for each studied alloy are equivalent. Hysteresis is always present to some extent in physical samples, but the minimal separation between the plateau pressures is a good indication that all three alloy powders exhibit fully reversible Mg-MgH 2 reaction at 300°C.
The pressure curves previously reported by Ishikawa et al. [40] for Mg 85 Zn 6 Y 9 powders show a significant pressure hysteresis between absorption and desorption (3.0 atm vs. 1.3 atm) at 300°C. The hysteresis decreased with increasing temperature until 400°C where the absorption and desorption branches of the diagram converged. In comparison, the alloy powders in this study show no hysteresis and achieve equivalent plateaus pressures at 300°C, which is attributed to the reduced crystallite size improving desorption kinetics. This improved kinetics enables reaching true Mg-MgH 2 equilibrium during experimentally accessible times. Ishikawa et al. [40] stated that the difference in their absorption and desorption pressures meant that the hydrogenation reaction in the Mg 85 Zn 6 Y 9 alloy is not completely reversible. However, the powder sample in the cited study was filed from the as-cast ingot and sieved through a mesh selecting particles \ 150 lm. While it cannot be confirmed if the pressure hysteresis observed by Ishikawa et al. [40] is the result of slow desorption kinetics due to large grains, it is both reasonable to assume that this is the case and that it could be mitigated based on the results of this study.
It should be noted that no significant difference was observed in the hydrogen plateau pressure of the Mg-Y-Zn alloys studied in the present work compared to pure MgH 2 , which has a plateau pressure of 1.8 atm at 300°C. Relative difference in plateau pressures can be used to qualitatively estimate changes in thermodynamics of hydride formation, with the similar plateau pressures indicating that MgH 2 decomposition thermodynamics in the Mg-Y-Zn powders is unaltered. This is not unusual as it is difficult to alter the energy of the metal-hydrogen bond by microstructure manipulations. The classic study by Huot et al. [5] on (de)hydrogenation of nanocrystalline MgH 2 established that HEBM did not alter the thermodynamics of the material and only allowed it to reach real equilibrium through the improved kinetics of the nanocrystalline structure. Another recent study by Yang et al. [43] on Mg 24 Y 3 M 1 (M = Ni, Co, Cu, Al) alloys with LPSO structures found that the substitution of M had no effect on MgH 2 absorption or desorption pressures at 320°C, showing that the thermodynamics are unaffected by varying these metals.

Microstructure of decomposed LPSO structures in hydrogenated thin Mg-Y-Zn foils
In nearly every study on hydrogenation of LPSO phase-containing alloys published to date, hydrogeninduced LPSO phase decomposition has been investigated in alloys processed into powders through manual filing or ball-milling. The small size of the initial particles and of the decomposed nanocomposite makes it difficult to observe the microstructure evolution of LPSO phase upon hydrogenation. The microstructural change that initiates the decomposition and the role of filing and ball-milling is an ongoing question. The high affinity for hydrogen by rare-earth elements leads to their hydrides forming early in the absorption process, and it is thought that the strain caused by this phase nucleation provides the driving force for decomposition. It may also be that formation of rare-earth hydrides destabilizes the LPSO, as it depends on the stabilizing effect of the rare-earth atoms.
To gain further insight into the LPSO decomposition process, the thin ZE3 foils prepared as described in Sect. ''Materials preparation'' were hydrogenated under different hydrogen partial pressures. The hydrogenated foils were ground and polished for SEM imaging, with the preparation process exposing microstructures at different depths in the foils due to the rough surface of the foils after hydrogenation. A variety of microstructures were thus seen throughout the foils below the surface.
The microstructural variations observed in the partially hydrogenated ZE3-P foil are seen in Fig. 11. Figures 11a, b, c show one type of decomposition microstructure observed close to the foil surface, where an extensive network of cuboid particles is embedded in the Mg/MgH 2 matrix. None of the initial LPSO structure is observed nor are any phases containing Zn. The composition of the particles has not been analysed, but EDS measurements from similar particles in the fully hydrogenated ZE3-F foil finds that they are nearly entirely yttrium with no significant amounts of Mg or Zn [56]. EDS cannot detect hydrogen, but as yttrium hydrides are expected to form it is concluded that the particles are YH 2 . The average size of the particles is between 1 lm and 2 lm and has a range from the submicron scale up to 5 lm. Recently, the GdH 2 particles of similar size were observed during hydrogenation studies of Mg (Gd) solid solutions [68]. 11d, e, f in the foil is unknown, but it is estimated to be within 30 lm of the surface due to the penetration of hydrogen into magnesium by diffusion [4]. EDS measurements in Figs. 11g, h show that the dark plates in the lamellar structures are Mg phase, and the lighter plates are a Mg-Zn phase. The lamellar structures are varied in domain size and orientation, with domain sizes ranging from the submicron to hundreds of microns. Variations in length, width, and orientation of the lamellae are attributed to the local orientation of the LPSO phase that was present before hydrogen decomposition. The cuboid YH 2 particles are also present but in small numbers without the clustering seen in Figs. 11a, b, c. The relative absence of YH 2 particles in the lamellar structures indicates that yttrium has segregated out of the LPSO phase.  Variations in the decomposition microstructures are also seen in the fully hydrogenated ZE3-F foil, seen in Fig. 12. Just below the oxidized surface layer of the foil in Figs. 12a, b, c, the microstructure consists of large Mg grains with a thin secondary phase located at the Mg grain boundaries. Higher magnification of this secondary phase in Fig. 12c reveals that it is an ultrafine lamellar Mg/Mg-Zn structure like the structure seen in the ZE3-P foil. This shows that hydrogen absorption has induced the same LPSO decomposition behaviour in both foils, regardless of the degree of hydrogenation.
At greater depths in the ZE3-F foil, the microstructures in Figs. 12d, e, f show that LPSO decomposition is incomplete. This incomplete decomposition indicates that the depth of the structures of Figs. 12d, e, f are within 30 lm of the surface of the foil, as hydrogen has clearly diffused deep enough to begin the decomposition reaction. The LPSO phase that has decomposed forms the Mg/Mg-Zn lamellar structures along the grain boundaries of Mg phase. Y-rich particles are seen within the Mg grains and at the boundaries between Mg grains and the lamellar Mg/Mg-Zn structures. The Mg grains at this depth are smaller compared to the grains seen near the surface of the foil, which is attributed to LPSO decomposition initiating at the surface as hydrogen is absorbed as well as enhanced heat transfer that promoted grain growth. EDS measurements in Figs. 12g, h find that the phase compositions of the Mg phase and the dark and light lamellae in the ZE3-F foil are the same as the ZE3-P foil. The appearance of Mg phase in both the ZE3-P and ZE3-F foils shows that LPSO decomposition will occur at hydrogen partial pressures lower than required to form MgH 2 . This observation along with the segregated YH 2 particles shows the necessity of the rareearth atoms to stabilize LPSO structures. Figure 13 shows a region of the ZE3-P foil where the cuboid particles are found embedded in the lamellar structure. There is no apparent orientation relationship, but particles observed in profile are flat rectangular prisms with depths that are the same size as the Mg-Zn lamellae. The size and position of the particles show that YH 2 nucleates within the LPSO phase, with the growth influenced by the orientation of the stacking faults. Future studies should determine which crystal directions along the plane of the stacking fault favour hydrogen diffusion and hydride particle growth.

Summary and conclusions
The effect of LPSO phase volume fraction on the hydrogen absorption and desorption properties of extruded Mg-Y-Zn alloys has been investigated. The outcomes of the research are: 1. Alloys with different LPSO phase volume fraction have been successfully prepared by control of solute content and casting, heat treatment, and extrusion conditions. The trend of increasing LPSO phase fraction with increasing solute concentration is confirmed with the prepared alloys. 2. Comparison of the LPSO phase volume fractions between the as-cast, heat-treated, and extruded alloys indicated that the medium solute content Mg 92 Y 5 Zn 3 alloy is most sensitive to thermal and mechanical processing, as the LPSO phase volume fraction significantly decreased after both heat treatment and extrusion. In contrast, total LPSO phase fraction in the alloys with low-and high solute content is relatively stable after the same treatments.

The hydrogen absorption kinetics increase in the
Mg-Y-Zn alloys with increasing concentration of alloying elements and the associated increase of the LPSO phase volume fraction. The ZE3 alloy with 99% LPSO phase volume fraction exhibits the highest absorption rate, closely followed by the ZE2 alloy with 58% LPSO phase volume fraction.  transition metal in these alloys. Compared to the equilibrium pressure of MgH 2 , all of the alloys have equivalent or lower equilibrium pressures, which indicates that MgH 2 has not been destabilized. The lack of a pressure hysteresis is attributed to the increased surface area of ballmilled powders accelerating the kinetics of equilibration and enabling reaching true thermodynamic Mg-MgH 2 equilibrium during experimentally accessible times. 6. To exclude the effect of ball-milling on LPSO phase decomposition, thin foils of high-content Mg-Y-Zn alloy consisting almost entirely of LPSO phases were hydrogenated under two different partial hydrogen pressures below and above the equilibrium Mg-MgH 2 pressure. Both partially and fully hydrogenated foils had similar microstructures and phases indicating full decomposition of the LPSO phase. This shows LPSO structure decomposition can occur at hydrogen pressures below those that form MgH 2 , emphasizing the critical role of YH 2 formation in LPSO phase destabilization.
The data set and results that support the findings of this study are openly available in https://hdl.handle. net/10779/DRO/DU:21366234.v1.

Declaration
Conflict of interest The authors declare that they have no conflict of interest or completing interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licen ses/by/4.0/.