2019 Volume 60 Issue 5 Pages 642-651
Microstructure formation mechanism through competitive reactions during initial hydrogenation in Mg/Cu super-laminate composites (SLCs) was investigated. Mg/Cu SLCs (Mg2Cu composition) were fabricated by accumulative roll bonding (ARB) and composed of laminate structures of Mg and Cu layers with the thickness of few hundreds nm. During the heating process of initial hydrogenation of Mg/Cu SLCs, hydrogenation of Mg and alloying of Mg with Cu followed by hydrogenation of Mg2Cu occurs competitively. It is found that microstructures of Mg/Cu SLCs during initial hydrogenation have changed drastically depending on the order of hydrogenation of Mg and Mg2Cu. The microstructures of Mg/Cu SLCs after initial hydrogenation can be categorized in three types such as (1) MgCu2 network, (2) MgCu2 sheath and (3) MgCu2 layer. Features of differential scanning calorimetry (DSC) profiles of the first cycle were well explained by this microstructure formation mechanism. In order to achieve only MgCu2 network structure, it is important to get fine, even and uniform microstructures in Mg/Cu SLCs. The large number of ARB cycles is inefficient. Changing flow properties such as annealing during ARB, warm-rolling and ultrasonic assisted rolling can be good strategies for that purpose.
It is crucial to solve the problem of global warming in order to build sustainable society. Low-emissions of CO2 are one of the most important issues for global warning countermeasures. It is effective to use renewable clean energies instead of fossil fuels in order to cut CO2 emissions. Hydrogen energy is the one of the most promising clean energies. It is essential to establish the new technologies to store hydrogen in a safe and easy to use form in order to utilize hydrogen as clean energies. It is strongly desired to develop hydrogen storage materials which can store hydrogen with high capacity.
On the one hand magnesium is still considered as a good candidate for hydrogen storage materials thanks to its high hydrogen storage capacity of 7.6 mass% in the form of reversible MgH2,1,2) low cost, light weight and high abundance. However, on the other hand its hydrogen absorption/desorption kinetics is slow even at elevated temperatures (>573 K) associated with its strong thermodynamic stability of standard formation enthalpy −74 kJ (mol H2)−1 and this results in limitations of its use.3) In order to solve these drawbacks, many efforts have been made in last few decades, such as appropriate alloying of Mg,4–6) adding catalysts7–11) and manufacturing nano-crystalline powder.12,13) Especially, promising results have been obtained with Mg–Ni based hydrogen storage alloys prepared by mechanical alloying.14–16) However, this synthesis process is rather expensive and not easily scalable.17) Therefore, it is important to put endeavors on the establishment of new and more efficient methods of producing metal hydrides. From this viewpoint, cold-rolling appears as an interesting alternative method because this process is simple, low cost, well known in the industry and easily scalable.
Ueda et al.18) have shown that accumulative roll bonding (ARB) is effective for the preparation and hydrogenation of Mg–Ni based hydrogen storage alloys. Since then, super-laminate composites (SLCs) prepared by ARB have been intensively investigated on the hydrogenation properties of Mg–Cu,19,20) Mg–Pd,21–23) Mg–Al,24) and Mg–Ti–Ni25) systems.
For instance, Mg/Cu SLCs exhibit reversible hydrogen absorption/desorption at 473 K after initial activation under the condition at 573 K in H2 atmosphere of 3.3 MPa,19) whereas Mg2Cu alloys prepared by conventional melting method require 573 K.26) From our previous investigation,20) it is known that the dynamics of hydrogen absorption/desorption is improved because the large surface area and the short diffusion distance are achieved after the initial activation of Mg/Cu SLCs at 573 K and this results in lowering the hydrogen absorption/desorption temperature to 473 K in equilibrium.
In initial activation, hydrogen is introduced at room temperature (R. T.) preventing from oxidation of Mg. Hydrogen absorption starts about at 453 K during the heating process of initial activation of Mg/Cu SLCs (hydrogenation under the condition at 573 K in H2 atmosphere of 3.3 MPa and dehydrogenation at 573 K in vacuum). At the early stage of microstructure formation during the heating process of initial activation, hydrogenation of Mg and alloying of Mg with Cu followed by hydrogenation of Mg2Cu occur competitively27–29) because Mg/Cu SLCs fabricated by ARB are composed of laminate structures of Mg and Cu layers with the thickness of few hundreds nm as seen in Fig. 1(a). This causes the multipath reactions forming complicated microstructures. It should be noted that Mg2Cu is formed at the interface between Mg and Cu layers during ARB as seen in Fig. 1(b).
Reflected electron images by SEM of as-rolled and typical three types of microstructures in Mg/Cu SLCs after initial hydrogenation: (a), (b) as-rolled, (c) MgCu2 network type, (d) MgCu2 sheath type and (e) MgCu2 layer type.
Inexplicably, Mg/Cu SLCs shows unstable initial hydrogenation properties and a variety of microstructures after initial hydrogenation although these are fabricated under the same condition. It is considered to be due to the quality of being uneven and lacking uniformity of Mg/Cu SLCs.
Figures 1(c), (d) and (e) show reflected electron images by a scanning electron microscope (SEM) of typical three types of microstructures in Mg/Cu SLCs after initial hydrogenation. They were hydrogenated under the conditions at 573 K for 86.4 ks (24 h) in H2 atmosphere of 3.3 MPa with a Sieverts’ type instrument. Figure 1(c) is MgCu2 network type that is usually seen in Mg/Cu SLCs which have fine as-rolled structures with more than 20 ARB cycles, (d) MgCu2 sheath type that is often seen in those which have medium as-rolled structures with 10 ARB cycles and (e) MgCu2 layer type that is mainly seen in those which have coarse as-rolled structures with 5 ARB cycles.
Figure 2 also shows three types of differential scanning calorimetry (DSC) profiles of first and second cycles. Hydrogen of 3.1 MPa was introduced at R. T. The heating/cooling rate was 5 K/min. Judging from the estimation of hydrogen absorption/desorption equilibrium temperature,29) the couples of endothermic and exothermic peaks around 720 K account for the hydrogen absorption/desorption peaks of Mg and those around 650 K account for those of Mg2Cu.
Three types of DSC profiles.
The features of these are summarised as below.
Figure 2(a): there are a very gentle exothermic peak around 520 K, a hydrogen desorption peak of Mg2Cu around 680 K and that of Mg around 730 K during the heating process of the first cycle. A hydrogen desorption peak of Mg2Cu of the second cycle is higher than that of the first cycle and that of Mg of the second cycle is lower than that of the first cycle. It is also seen that a hydrogen absorption peak of Mg around 710 K and that of Mg2Cu around 630 K during the cooling process of the first cycle. The hydrogen absorption peak of Mg during the cooling process of the first cycle is small and barely seen during that of the second cycle. The hydrogen absorption peak of Mg2Cu in the second cycle is higher than that in the first cycle.
Figure 2(b): there is a very gentle exothermic peak around 650 K. The hydrogen desorption peak of Mg2Cu is slightly seen and that of Mg around 730 K is the higher than that in Fig. 2(a) during the heating process of the first cycle. A hydrogen desorption peak of Mg2Cu of the second cycle is clear and high, and that of Mg of the second cycle is lower than that of the first cycle. It is also seen that a hydrogen absorption peak of Mg around 710 K, which is higher than that in Fig. 2(a), and that of Mg2Cu around 630 K, which shows a long tail toward lower temperature, during the cooling process of the first cycle. The hydrogen absorption peak of Mg during the cooling process of the first cycle is the almost same height as that of the second cycle. The hydrogen absorption peak of Mg2Cu in the second cycle is higher than that in the first cycle and shows a long tail, too.
Figure 2(c): there is an extremely gentle exothermic peak around 520 K. The hydrogen desorption peak of Mg2Cu is barely seen and that of Mg around 730 K is the highest among Fig. 2(a), (b) and (c) during the heating process of the first cycle. The hydrogen desorption peak of Mg2Cu in the second cycle shows a double peak and that of Mg in the second cycle is lower than that in the first cycle. It is also seen that a hydrogen absorption peak of Mg around 710 K, which is the highest among Fig. 2(a), (b) and (c), and that of Mg2Cu around 630 K, which does not shows a long tail toward lower temperature during the cooling process of the first cycle like Fig. 1(b). The hydrogen absorption peak of Mg during the cooling process of the first cycle is the almost as high as that of the second cycle. The hydrogen absorption peak of Mg2Cu also shows the same tendency as that of Mg.
The features mentioned above still appeared even after twenty cycles of DSC measurements.19) This fact indicates that the initial structures of Mg/Cu SLCs play an important role in hydrogen absorption/desorption properties.
It is considered that the variety of microstructures after initial hydrogenation and hydrogen absorption/desorption properties of Mg/Cu SLCs originates from the quality of being uneven and lacking uniformity of Mg/Cu SLCs. However, the detail of them is not fully understood yet. Therefore, it is important to clarify why these differences of microstructures and properties can occur in order to embody the uniform and superior properties. In this paper, we examined the microstructure formation mechanism through competitive reactions during initial hydrogenation in Mg/Cu SLCs and the relationship between microstructures and hydrogen absorption/desorption properties. There, we clarified the order and temperature range of competitive reactions during initial hydrogenation and revealed the physical meaning of ARB from a microstructural point of view. We also suggested the point that should be reconsidered about the fabrication process of Mg/Cu SLCs, although only the relationship between microstructures and hydrogen absorption/desorption properties was mentioned in previous papers.19,20,27–29)
ARB was used to get Mg/Cu SLCs. We used Commercial pure Mg (Takeuchi Metal Foil & Powder Co., LTD., 99.9% of purity, 250 µm in thickness) and oxygen-free Cu (Fukuda Metal Foil & Powder Co., LTD, 99.98% of purity, 10 µm in thickness) as starting materials. Mg foils were polished with #600 sandpaper in order to remove oxide layers and cold-rolled to 40 µm. These 40 µm Mg foils and 10 µm Cu foils were cut into 20 mm × 40 mm in size. After that, they were annealed in order to remove internal stress and strain in Ar atmosphere of 0.1 MPa for 3.6 ks, at 673 K for Mg and at 873 K for Cu respectively. The surfaces of both kinds of foils were cleaned with dilute hydrochloric for Mg and nitric acid for Cu respectively. Twenty pairs of Mg foils with 40 µm in thickness and Cu foils with 10 µm in thickness were stacked in one as in the stoichiometric composition of Mg2Cu. This stack was wrapped up in a Cu plate with 1 mm in thickness and pressed under the condition of 735 MPa for 300 s prior to ARB in order to enhance adhesion and to prevent slip between Mg and Cu foils during ARB. The stack was cold-rolled in a conventional roll-bonding process at room temperature in air. Then, the length of rolled materials is sectioned into two halves. The sectioned foils were stacked one on top of the other, wrapped up in a Cu plate and cold-rolled again. This procedure was repeated to fabricate a Mg/Cu SLC. In effect, ARB rolling is not only a deformation method but also a bonding process. The process can introduce ultra-high plastic strain without any geometric change.30)
It is important to know that the half of surface regions comes to the center in the next cycle in ARB process. This results in complicated distributions of the surface regions with large shear strain through thickness of the foil. In order to achieve different initial (as-rolled) structures, four types of specimens were prepared by changing the number of above procedure, which are 5, 10, 20 and 30 cycles of ARB.
In order to know the hydrogen absorption/desorption properties depending on the number of ARB, Mg/Cu SLCs were hydrogenated under the conditions at 573 K for 86.4 ks in H2 atmosphere of 3.3 MPa with a Sieverts’ type instrument. DSC measurements were also performed.
From the experiments of Mg–Cu diffusion couples,31,32) it is known that Mg2Cu and MgCu2 can be formed at the interface between Mg and Cu layers in Mg/Cu SLCs at elevated temperature19,29) so that Mg2Cu is at Mg side and MgCu2 is at Cu side. In order to know the growth rate constants of Mg2Cu in Mg/Cu SLCs with different ARB cycles, diffusion annealing of Mg/Cu SLCs was performed. Since it is pointed out that MgCu2 is not formed under 473 K33,34) and known from our previous experiments that Mg/Cu SLCs starts to absorb hydrogen around at 453 K, diffusion annealing was performed under 453 K. The growth length of Mg2Cu layers during diffusion annealing was estimated by combining SEM observations and integrated intensity ratio analyses of X-ray diffraction (XRD). The measurement method of the growth rate constants of Mg2Cu in Mg/Cu SLCs was reported in detail in Ref. 35 and the experimental results of the growth rate constants of those with different cycles of ARB will be reported in detail elsewhere.
In order to estimate the formation order of phases formed during heating process (alloying process), small pieces of Mg/Cu SLCs were hated up at every 50 K between 423 K and 773 K with the heating rate of 10 K/min and cooled by turning off the heater in DSC instrument in Ar atmosphere of 3.1 MPa. Constituent phases were confirmed by a powder XRD and energy dispersive X-ray spectroscopy (EDS). Mg/Cu SLCs were crushed for XRD in order to avoid the influence of a texture developed during ARB.
In order to know the temperature range of multipath reactions during initial hydrogenation for the Mg/Cu SLCs with 20 ARB, DSC measurements were performed on small pieces of Mg/Cu SLCs at the temperature range between 373 K and 753 K with the heating and cooling rate of 10 K/min in Ar and H2 atmosphere of 3.1 MPa. DSC measurements in Ar atmosphere is performed to estimate the reactions during heating process (alloying process) before hydrogenation and the subtraction of DSC measurement in Ar atmosphere from that in H2 atmosphere was evaluated to know the reactions during hydrogenation process. DSC profiles were peak-fitted by pseudo-Voigt function after subtracting back grounds.
It is expected in this study that DSC profiles contain many peaks caused by the multipath reactions mentioned above. In order to know the information about temperature range of each constituent reaction in these multipath ways, peak separation was tried. Many parameters (activation energy, growth rate constant, conversion ratio and more) are necessary for each reaction to fit profiles based on complicated theoretical approach to the experimental ones, in spite that there are insufficient data for those parameters. This time we applied simple pseudo-Voigt function composed of Gaussian and Lorentzian functions for the profile-fitting, apart from the theoretical approach for chemical reaction, in order to know the temperature range roughly.
Microstructures of Mg/Cu SLCs were observed with SEM and a scanning transmission electron microscope (STEM). SEM and STEM specimens were prepared by conventional mechanical grinding and polishing followed by ion milling.
The formation mechanism of microstructures in Mg/Cu SLCs during initial hydrogenation through competitive reactions was estimated by evaluating the growth rate constants, XRD profiles, DSC profiles and microstructures of Mg/Cu SLCs.
Concerning the initial hydrogenation properties of as-rolled Mg/Cu SLCs with different ARB cycles under the conditions at 573 K for 86.4 ks (24 h) in H2 atmosphere of 3.3 MPa, Mg/Cu SLCs with 5 and 10 cycles of ARB absorb hydrogen slowly, whereas those with 20 and 30 cycles of ARB absorb hydrogen fast.36) From SEM observations of those specimens, it can be said that the total area of MgCu2 network were 30 cycles > 20 cycles > 10 cycles > 5 cycles. It seems that there are some differences between over 20 cycles and under 10 cycles.
Figure 3 shows DSC profiles of Mg/Cu SLCs with different ARB cycles. Hydrogen of 3.1 MPa was introduced at R. T., the heating rate from R. T. to 753 K was 20 K/min and cooling rate from 753 K to 373 K was 10 K/min. Only hydrogen desorption peak of Mg is seen in Fig. 3(a) and (b) with 5 and 10 cycles of ARB during heating process, whereas hydrogen desorption peak of Mg2Cu can be seen in Fig. 3(c) and (d) with 20 and 30 cycles of ARB during heating process. This indicates enough amount of Mg2Cu did not form in Mg/Cu SLCs with 5 and 10 cycles of ARB during heating process. Hydrogen absorption peak of Mg2Cu cannot be seen in Fig. 3(a), is low in Fig. 3(b) and high in Fig. 3(c) and (d). It also seems that there are some differences between over 20 cycles and under 10 cycles. From these results, it can be said that hydrogenation of Mg2Cu is dominant in Mg/Cu SLCs with 20 and 30 ARB cycles and the hydrogenation of Mg is dominant in Mg/Cu SLCs with 5 and 10 ARB cycles.
DSC profiles of Mg/Cu SLCs with different ARB cycles: (a) 5 cycles of ARB, (b) 10 cycles, (c) 20 cycles and (d) 30 cycles.
The growth rate constants of Mg2Cu in Mg/Cu SLCs with different ARB cycles were examined35) because we expected that the growth rate constants of Mg2Cu in Mg/Cu SLCs increased with the number of ARB cycles. Although the growth rate constants of Mg/Cu SLCs with 5 cycles of ARB could not be measured because Mg2Cu did not grow enough owing to the insufficient contact between Mg and Cu foils, it is found from results of diffusion annealing for Mg/Cu SLCs with 10, 20 and 30 cycles of ARB that the growth rate constants k for the Mg2Cu layers can be described by an Arrhenius equation as below:31,32,35)
| \begin{equation*} k = k_{0}\exp(-Q/RT) \end{equation*} |
It is seen that the activation energies for the layer growth process of Mg2Cu layers in Mg/Cu SLCs are much lower than those of Mg–Cu diffusion couples.31,32) As a result, the growth rate constants of Mg2Cu in Mg/Cu SLCs are approximately 102–104 larger than that of Mg–Cu diffusion couples below 453 K. It indicates that Mg2Cu can grow at low temperature with sufficient rate in Mg/Cu SLCs. It can be said that the growth rate constants were increased by ARB. However, there was no big difference among Mg/Cu SLCs with different ARB cycles contrary to our expectation.
It is reported that severe plastic deformation (SPD) such as high-pressure torsion (HPT) and equal-channel angular pressing (ECAP) can enhance atomic diffusion and promote solid-state reactions because of an increase in the density of lattice defects such as vacancies, dislocations and grain boundaries.35,37–41) It is reasonable to consider the introduction of a high density of vacancies and dislocations, and the grain refinement during ARB process.
The introduction of high density dislocations and grain refinement in as-rolled Mg/Cu SLCs with 5 and 10 cycles of ARB were confirmed by STEM observations in Ref. 36. The average grain size of Mg in the case of 5 cycles of ARB, 590 nm was larger than that of 20 cycles of ARB, 180 nm. (It should be noted that an average grain size is different from an average layer thickness.) However, that is because of inhomogeneity. The grain size of most grains in Mg/Cu SLCs with 5 cycles of ARB was almost as same as 20 cycles of ARB, although large grains in Mg/Cu SLCs with 5 cycles of ARB made the average grain size lager. The dislocation density of Mg and Cu was high in both specimens. As seen in Fig. 5(b) (5 cycles of ARB) and (d) (20 cycles of ARB) of Ref. 36, the same things can be said about Cu. That is the reason why the growth rate constants were almost same among Mg/Cu SLCs with 10, 20 and 30 cycles of ARB.
Figure 4 shows the simulation of dependency on time and temperature of Mg2Cu volume ratio in Mg/Cu SLCs with different ARB cycles. The values of thickness of Mg and Cu, k0 and Q listed in Table 1 are used. The feature of Fig. 4(b) 20 ARB cycles and (c) 30 ARB cycles are almost same although that of Fig. 4(a) 10 cycles is different from these. It can be said that the volume ratio of Mg2Cu at the same time and same temperature become higher as the number of ARB cycles increases. Because the number of interface between Mg and Cu layers where Mg2Cu grows increases and the thickness of Mg decreases as the number of ARB cycles increases. Figure 5 shows the thickness of Mg in Mg/Cu SLCs as a function of the number of ARB cycles. It is clearly seen that the thickness of Mg decreases drastically above 5 ARB cycles and that is almost same at 20 and 30 ARB cycles. Since the growth rate constants were almost same among Mg/Cu SLCs with 10, 20 and 30 ARB cycles, the growth length of Mg2Cu should be almost same among them. If Mg is thick, Mg cannot transform to Mg2Cu completely within a certain time at a certain temperature. That is the reason why the hydrogenation of Mg2Cu is dominant in Mg/Cu SLCs with 20 and 30 ARB cycles and the hydrogenation of Mg is dominant in Mg/Cu SLCs with 5 and 10 ARB cycles although the growth rate constants are almost same among Mg/Cu SLCs with 10, 20 and 30 ARB cycles.
Simulation of dependency on time and temperature of Mg2Cu volume ratio in Mg/Cu SLCs with different ARB cycles: (a) 10 cycles of ARB, (b) 20 cycles and (c) 30 cycles.
Thickness of Mg in Mg/Cu SLCs as a function of the number of ARB cycles.
Figure 6(a) shows a DSC profile of Mg/Cu SLCs heated up to 753 K at the heating rate of 10 K/min in Ar atmosphere of 3.1 MPa. The back ground is subtracted already. Since exothermic peaks in Fig. 6(a) can be seen only in the heating process of the first cycle and cannot be seen in after the second cycle, it is considered that these peaks are related to the alloying process of Mg with Cu. Four peaks were resolved by peak-fitting using pseudo-Voigt function. Figure 6(b) shows a DSC profile of subtraction of Fig. 6(a) from a DSC profile of Mg/Cu SLCs heated up to 753 K at the heating rate of 10 K/min in H2 atmosphere of 3.1 MPa. Also, two peaks were resolved by peak-fitting using pseudo-Voigt function. From SEM and STEM observations, XRD and DSC measurements, each peak in DSC profiles was identified to each reaction as below.42)
| \begin{equation} \text{Peak 1:}\ \text{recovery and recrystallization of Cu} \end{equation} | (1) |
| \begin{equation} \text{Peak 2:}\ \text{formation of Mg$_{2}$Cu as 2Mg} + \text{Cu}\rightarrow\text{Mg$_{2}$Cu} \end{equation} | (2) |
| \begin{align} \text{Peak 3:}\ &\text{formation of MgCu$_{2}$ as Mg} + \text{2Cu}\rightarrow\text{MgCu$_{2}$}\\ &\text{or Mg$_{2}$Cu} + \text{3Cu}\rightarrow\text{2MgCu$_{2}$} \end{align} | (3) |
| \begin{align} \text{Peak 4:}\ &\text{formation of Mg$_{2}$Cu as 3Mg} + \text{MgCu$_{2}$}\\ &\rightarrow\text{2Mg$_{2}$Cu} \end{align} | (4) |
| \begin{equation} \text{Peak 5:}\ \text{hydrogenation of Mg as Mg} + \text{H$_{2}$}\rightarrow\text{MgH$_{2}$} \end{equation} | (5) |
| \begin{align} \text{Peak 6:}\ &\text{hydrogenation of Mg$_{2}$Cu as 2Mg$_{2}$Cu} + \text{3H$_{2}$}\\ &\rightarrow\text{3MgH$_{2}$} + \text{MgCu$_{2}$} \end{align} | (6) |
DSC profile of Mg/Cu SLCs: (a) heated up to 753 K at the heating rate of 10 K/min in Ar atmosphere of 3.1 MPa, (b) subtraction of (a) from a DSC profile of Mg/Cu SLCs heated up to 753 K at the heating rate of 10 K/min in H2 atmosphere of 3.1 MPa.
Figure 7 is a schematic diagram of these multipath reactions representing the temperature range and intensity of reactions. It includes dehydrogenation reactions reflecting Fig. 2. The thickness of horizontal bar indicates the intensity of reactions qualitatively. From Fig. 7, it can be understood easily that the formation of Mg2Cu (alloying of Mg with Cu) (2) followed by the hydrogenation of Mg2Cu (6) and the hydrogenation of Mg (5) are competitive at the early stage of microstructure formation during the heating process of initial activation of Mg/Cu SLCs. From the experiments of Mg–Cu diffusion couples,31,32) it is expected that core-shell like structures can be formed during heating process.19,29) Constituent phase of the core-shell like structures are Cu, MgCu2, Mg2Cu, and Mg in the order from the core to the shell.
Schematic diagram of multipath reactions representing the temperature range and intensity of reactions.
The formation mechanism can be classified in three cases depending on the order of hydrogenation of Mg and Mg2Cu as below.
Figure 8 shows the schematic models of microstructure formation mechanism through competitive reactions during initial hydrogenation of Mg/Cu SLCs. If we consider the whole of core-shell like structure, such as existence of Cu or MgCu2 in the core, there can be many kinds of structures. However, the inside of core-shell like structure is not essential. Therefore, we consider typical three cases as seen in Fig. 1(c), (d) and (e).
Schematic models of microstructure formation mechanism through competitive reactions during initial hydrogenation of Mg/Cu SLCs. Case I: MgCu2 network type, Case II: MgCu2 sheath type, Case III: MgCu2 layer type.
In case I, the formation of Mg2Cu starts first. Then Mg2Cu grows and is hydrogenated so that Mg2Cu shell disproportionates to MgH2 + MgCu2 to form the open MgCu2 network coexisting with MgH2 after initial hydrogenation as seen in Fig. 1(c). This situation is the same as hydrogenation of Mg2Cu single phase.27)
In case II, the formation of Mg2Cu starts first. Then Mg2Cu grows. However, the hydrogenation of Mg occurs before that of Mg2Cu. When Mg2Cu is hydrogenated, MgH2 forms along the surrounding MgH2 that is hydrogenated previously. Since it is expected that the surfaces of MgH2 surrounding Mg2Cu can be nucleation sites of MgH2 formed in the process of hydrogenation of Mg2Cu to disproportionate to MgH2 + MgCu2, the MgCu2 can grow along the surface of Mg2Cu in a way to cover Mg2Cu. As a result, the MgCu2 sheath is formed as seen in Fig. 1(d). It should be noted that hydrogenation of Mg2Cu seems to have stopped when it is surrounded by MgCu2 completely as seen in Fig. 1(d). This is not contradictory to that Mg/Cu SLCs with 10 cycles of ARB absorb hydrogen slowly, whereas those with 20 and 30 cycles of ARB absorb hydrogen fast.36) Similar situation occurs during cooling process of DSC measurements as seen in Fig. 2(b). During this, the hydrogenation of Mg occurs first around 710 K and then that of Mg2Cu occurs around 630 K. The long tail toward lower temperature in hydrogenation peak of Mg2Cu can be understood to be due to the MgCu2 sheaths. In Fig. 2(c), this tendency is weak in the second cycle of DSC measurement. Because the situation is different in Fig. 2(c) from Fig. 2(b) since case III is considered to occur during the first cycle of DSC measurement as explained as below.
In case III, the hydrogenation of Mg starts first. Then MgH2 reacts with Cu following the equation that MgH2 + 2Cu → MgCu2 + H2 to form the MgCu2 layer as seen in Fig. 1(e). This reaction is dehydrogenation of MgCu2. It should be noted that pores often stay adjacent to the MgCu2 layers in Fig. 1(e). It is considered to be due to the hydrogen desorption. The existence of this reaction is confirmed in previous experiments,27,28) although the reaction of MgH2 with Cu under H2 atmosphere is not known in general. Since MgCu2 does not absorb hydrogen under moderate conditions, case III should be avoided. It is considered to be important that Mg2Cu is formed between Mg and Cu during ARB for the growth of Mg2Cu at low temperature.
From Fig. 4, it is known that Mg2Cu volume ratio in Mg/Cu SLCs can be controlled by the reaction temperature and thickness of Mg + Cu. It is also important that Mg is thin so that the hydrogenation of Mg2Cu is dominant. From the view point of handling, it is not adequate to use much thinner Mg and Cu foils since Mg and Cu foils with the thickness of 40 and 10 µm are used originally as starting materials. Also, the large number of ARB cycles is inefficient since the thickness of Mg in Mg/Cu SLCs with 20 and 30 ARB cycles is almost same as seen in Fig. 5, although the large number of ARB cycles can improve the quality of being uneven and lacking uniformity of Mg/Cu SLCs judging from the microstructure of as-rolled Mg/Cu SLCs with 20 and 30 ARB cycles as seen in Fig. 9.
Reflected electron images by SEM of as-rolled Mg/Cu SLCs after (a) 5, (b) 10, (c) 20 and (d) 30 ARB cycles.
There are two methods to form Mg2Cu phase from Mg/Cu SLCs. One is to hydrogenate and dehydrogenate Mg/Cu SLCs. Another is to anneal them in inert gas atmosphere directly. On the one hand, higher temperature is more efficient in annealing method. However, MgCu2 can be formed at higher temperature as a drawback. On the other hand, there is a possibility to form Mg2Cu effectively using dehydrogenation process in hydrogenation and dehydrogenation method. Another good point of this method is that the grain size of Mg2Cu prepared by hydrogenation and dehydrogenation method is relatively small, approximately 1 µm due to hydrogen disproportionation desorption recombination (HDDR)43–45) and this can be effective in hydrogen absorption kinetics. Shao et al.12) pointed out that it was not easy to obtain pure Mg2Cu compound from metal nanoparticle in convenient conditions at 673 K in Ar atmosphere of 0.1 MPa for 25.2 ks (7 h) and the better result was obtained when Mg2Cu alloy was heated under H2 atmosphere of 4.0 MPa for 32.4 ks (9 h) and evacuated for 7.2 ks (2 h) at 673 K.
We did initial activation of Mg/Cu SLCs to obtain Mg2Cu (hydrogenation under the condition at 573 K, in H2 atmosphere of 3.3 MPa and dehydrogenation under the condition at 573 K in vacuum). Remained Cu did not disappear completely although hydrogenation and dehydrogenation repeated several times, whereas remained Cu disappear after 20 cycles DSC measurements. The higher temperature in hydrogenation and dehydrogenation method might be more effective.
The metallic multilayers produced by deformation process such as ARB can be divided into the following two categories. In one, the hard phase of the constituent layer necks and raptures into lamellae during the deformation process, which results in the hard lamellae phase embedded composite. In the other, the continuity of the constituent layers is maintained until the layer thickness reduced to nano-scale.46)
Figure 9 shows the microstructures of as-rolled Mg/Cu SLCs at RD-ND planes after various ARB cycles. As seen in Fig. 9(a), necking and rapture are observed obviously in Cu layers, leading to fragmentation of Cu layers in the Mg matrix after 5 ARB cycles. As the ARB proceeds, Cu fragments increase in the microstructures and disperse in the Mg matrix as seen in Fig. 9(b) to (d). Judging from Fig. 9, Mg/Cu SLCs seem to be categorized in the former case.
Necking and fracture take place in the hard phase due to simultaneous deformation of dissimilar metals and difference in their flow properties such as strength coefficient and strain hardening exponent and the initial thickness ratio of the constituent phases.47) When rapture starts to take place, fragment-free zone (surrounded by white lines) are formed in the microstructures. As seen in Fig. 9(b), the number and size of the fragment-free zone is reduced to some extent after 10 ARB cycles. However, the distribution of Cu fragments is not still uniform. Higher ARB cycles need to be applied to get a uniform distribution of Cu fragments such as Fig. 9(c) and (d).
In general, during the co-deformation of dissimilar metals plastic instability occur due to the differences in flow properties for the constituent phases and further deformation causes the hard phase to neck and finally rapture. Therefore, multilayer maintaining the layer continuity is more difficult to make. As explained above, three are three microstructure formation mechanisms originated from the quality of being uneven and lacking uniformity of Mg/Cu SLCs even after 20 or 30 ARB cycles during the heating process of initial hydrogenation. In order to make only the case I (MgCu2 network) take place, it is important to get fine, even and uniform microstructures in Mg/Cu SLCs. As explained about Fig. 5, more than 30 ARB cycles is inefficient for that purpose. Changing flow properties can be a good strategy. Therefore, it is considered that annealing during ARB, warm-rolling and ultrasonic assisted rolling are worth to try.
Microstructure formation mechanism through competitive reactions during initial hydrogenation in Mg/Cu SLCs was elucidated.
The formation of Mg2Cu (alloying of Mg with Cu) followed by the hydrogenation of Mg2Cu and the hydrogenation of Mg occurs competitively at early stage of microstructure formation during the heating process of initial activation of Mg/Cu SLCs.
The formation mechanism can be classified in three cases depending on the order of hydrogenation of Mg and Mg2Cu as below.
It is important to get fine, even and uniform microstructures in Mg/Cu SLCs to make only the case I (MgCu2 network) take place. Changing flow properties such as annealing during ARB, warm-rolling and ultrasonic assisted rolling can be good strategies for that purpose.
This work was supported by JSPS KAKENHI Grant number 23560794 and 15K06519. Authors appreciate Mr. K. Kurumatani, Mr. K. Ikeuchi, Mr. D. Nishino and Mr. K. Hayashi for assistance of experiments.
