III. RESULTS AND DISCUSSION

Figure 1 shows the nominal composition of the alloy studied relative to the phase fields determined in the literature; based on this, the key alloy composition chosen in this study falls within the Ca₂Mg₅Zn₅, Ca₂Mg₆Zn₃, and IM1 phase fields according to the respective Mg–Zn–Ca phase diagrams published by Paris, ^{Reference Paris18} Clark, ^{Reference Clark19} and Zhang. ^{Reference Zhang30} Newer Mg–Zn–Ca phase diagrams do exist, ^{Reference Brubaker and Liu33–Reference Wasiur-Rahman and Medraj35} but these were derived via computational optimizations based on the experimental data of Paris, Clark, and Zhang. Thus only those experimental phase diagrams are shown in Fig. 1. DSC analysis from 150 to 650 °C (Fig. 2) revealed three distinct exothermic peaks, which were expected to be, in decreasing order, the liquidus temperature, the eutectic temperature where the ternary phase–Mg lamellae begin to form, and the eutectoid transformation temperature of Mg₇Zn₃ → MgZn + Mg. ^{Reference Clark19,Reference Thaddeus, Massalski, Subramanian and Kacprzak36} The determined liquidus temperature of 417 °C is in line with the value compiled by Villars, ^{Reference Villars, Prince and Okamoto37} which is largely based on the work of Paris, ^{Reference Paris18} and is also similar to the 400 °C determined through thermal analysis during solidification of Mg–0.5Ca–xZn (x = 0.5, 1 and 3). ^{Reference Bakhsheshi-Rad, Abdul-Kadir, Idris and Farahany8} Most critically, the DSC analysis reveals that obvious phase transitions occur at temperatures above 335 °C. This observation supports past studies by Zhang ^{Reference Zhang, Kevorkov, Li, Essadiqi and Medraj28,Reference Zhang, Kevorkov, Bridier and Medraj38} and Clark, ^{Reference Clark19} which used diffusion couples annealed at 335 °C to explore the ternary Mg–Zn–Ca equilibrium phase diagram. At higher annealing temperatures, a liquid phase may exist (see eutectic temperature determined in Fig. 2) which would complicate the interpretation of the diffusion couple experiments. ^{Reference Kodentsov, Bastin and van Loo39}

FIG. 2.

Representative DSC cooling curve at a rate of 10 K/min for Mg_58.15Zn_29.45Ca_12.40, showing distinct exothermic peaks with onsets at 417, 369, and 332 °C. The inset illustrates that the peaks shown were the only exothermic events throughout the temperature range.

Microstructural analysis identified the phase transitions associated with each exothermic peak in the DSC trace. Using upper and lower temperatures of 600 °C and 335 °C, respectively, the gradient Bridgman method produced a rod sample with large polygonal crystals of between ∼100 and 500 µm in size that formed a quasibinary with the α-Mg matrix (Fig. 3). Thus the two high-temperature exothermic peaks in Fig. 2 were determined as the liquidus and eutectic temperatures. A selected area was subsequently mapped using EDX [Figs. 3(c)–3(e)], which confirmed that the large polygonal grains (bright phase) in the backscattered electron (BSE) image is the ternary intermetallic, and the dark phase in the BSE image is Mg (with Zn in solid solution). A number of additional grains were selected throughout the sample and analyzed by both SEM and TEM EDX to determine their compositions. All of these readings were consistent throughout the sample, and are tabulated including statistics in Table II. Despite the slightly higher Mg value in SEM EDX and the slightly higher Zn value in TEM EDX, these data show similar atomic concentrations of Mg and Zn, resulting in stoichiometric ratios of ∼Ca₂Mg₅Zn₅. We therefore refer to this phase as Ca₂Mg₅Zn₅. The Zn content in α-Mg is approximately 2.42 at.% (averaged from 5 SEM EDX readings), which is very close to the 2.4 at.% reported as the maximum solid solubility of Zn in Mg at 340 °C. ^{Reference Villars, Prince and Okamoto37}

FIG. 3.

(a) BSE image obtained from SEM. (b) Magnified area of image (a), on which elemental mapping was performed to investigate (c) Ca, (d) Mg, and (e) Zn.

TABLE II.

EDX SEM and EDX TEM elemental analysis of the grains within the Mg_58.15Zn_29.45Ca_12.40 sample prepared; n gives the number of measurements that generated a mean value with standard deviation (SD). All values are given in at.%.

It is worth emphasizing that no significant composition difference was observed across the individual grains. Although this work did not establish the solubility limits of Ca₂Mg₅Zn₅, work by Zhang ^{Reference Zhang, Kevorkov, Li, Essadiqi and Medraj28,Reference Zhang30,Reference Zhang, Kevorkov, Bridier and Medraj38} suggested that the Zn and Mg contents in IM1 Ca₃Mg _x Zn_15−x (4.6 ≤ x ≤ 12 at 335 °C) could each vary by up to 41.1 at.%, thus making Ca₂Mg₅Zn₅ a possible variation of IM1 (when x = 7.5, IM1 = Ca₃Mg_7.5Zn_7.5 or Ca₂Mg₅Zn₅). This observed difference was attributed to sample preparation methods, whereby the current method produced the equilibrium microstructure at 335 °C by controlled cooling of the alloy from its liquid state. The failure to obtain Ca₂Mg₅Zn₅ using diffusion couple experiments reaffirmed the known limitations of such methods, where the absence of certain phases may be due to difficulties in nucleation, or quasiequilibrated diffusion zones may have formed instead of the real equilibrium phase. ^{Reference Kodentsov, Bastin and van Loo39}

Microstructural analysis of the current study revealed large polygonal grains of Ca₂Mg₅Zn₅ and eutectic (Ca₂Mg₅Zn₅ + Mg) lamellae. According to Zhang, ^{Reference Zhang, Kevorkov, Li, Essadiqi and Medraj28} the same alloy composition would also generate a microstructure consisting of IM1 (encompassing Ca₂Mg₅Zn₅) + Mg. Distinctions must be made between these claims, and we therefore utilize the refined crystal structure parameters. Rietveld refinement of the XRD patterns for IM1 ^{Reference Zhang, Kevorkov, Li, Essadiqi and Medraj28} showed that this solid solution crystallizes in a hexagonal structure having a P6₃/mmc (194) space group with Sc₃Ni₁₁Si₄ prototype. These data should, however, be viewed with caution, as the suggested structure model had an increasingly worsening reliability factor, starting with R _wp = 15.8% (for IM1 = Ca₃Mg₁₂Zn₃) and reaching 20.5% (for IM1 = Ca₃Mg_7.8Zn_7.2) and 19.6% (for IM1 = Ca₃Mg_7.6Zn_7.4) ^{Reference Zhang, Kevorkov, Li, Essadiqi and Medraj28} ; the latter two IM1 compositions are very close to ∼Ca₂Mg₅Zn₅. Due to the poor structural fit, we thus carried out our own structure determination plus refinement work.

The crystal structure of Ca₂Mg₅Zn₅ was determined using single-crystal XRD methods. Six large grains of ∼100 µm in size were mechanically removed from the bulk sample and measured, and only the data from the highest quality grain was used in the later refinement steps. Ca₂Mg₅Zn₅ was found to be hexagonal with lattice parameters of a = 9.5949(3) Å, c = 10.0344(3) Å, and space group P6₃/mmc. Structure solution and refinement were solved from scratch with direct methods using the SHELX97 software package. ^{Reference Sheldrick40} Results are shown in Table III [further details of the crystal structure investigation(s) may be obtained from FIZ Karlsruhe, 76344 Eggenstein-Leopoldshafen, Germany (fax: (+49) 7247-808-666; e-mail: crysdata@fiz-karlsruhe.de, on quoting the deposition number CSD-429847)]. The solved structure can be described as an Sc₃Ni₁₁Si₄ type, as already reported in the literature, ^{Reference Zhang, Kevorkov, Li, Essadiqi and Medraj28} but with additional Zn atoms, Zn(5) (see Fig. 4), at the symmetry-related sites (0,0,0) and (0,0,0.5). While the past investigation used powder XRD ^{Reference Zhang, Kevorkov, Li, Essadiqi and Medraj28} with the assumption of a Sc₃Ni₁₁Si₄ prototype, the structure in this study was solved from single crystals with no prior assumptions. The initially ordered model showed unsatisfactory R-values and unusual large atomic displacement parameters (ADPs) for the Zn atoms, which indicated the presence of disorder. Because Mg and Zn may replace each other due to their similar bond distances, ^{Reference Senkov, Miracle, Barney, Hannon, Cheng and Ma41} the possibility of mixed Mg/Zn sites was considered for all Zn and Mg positions. The refinement results, however, clearly indicated the absence of any substitutional or occupational disorder for Zn(1) and Mg(6). For Zn(4) and Zn(5) the possibility of mixed occupied Mg/Zn sites was deemed unreasonable because their interatomic distance is only 2.51 Å, which is very short even for Zn–Zn pairs. Substitution of Zn by larger Mg atoms was therefore ruled out and only under-occupied Zn was examined for these sites. For Zn(2)/Mg(2) and Zn(3)/Mg(3) mixed parameters could be successfully modeled. The final disorder model clearly improved the fit: the SHELXL program specific R-values ^{Reference Sheldrick40} dropped from R1 = 9.0%, wR2 = 28.3% for the initial ordered model to R1 = 4.8%, wR2 = 9.7%. The ADPs of the Zn atoms are still relatively large, but this can be easily explained by size-effect distortions caused by substitutional or occupational disorder. The refined composition of the structure is Ca_16.0Mg_42.0Zn_42.0, which is very close to the findings obtained by EDX (Table II).

TABLE III.

List of (a) atomic coordinates and occupancies; and (b) ADPs from the single-crystal refinement of Ca₂Mg₅Zn₅, where ${U_{{\rm{eq}}}} = {1 \over 3}\left\langle {\mathop {\left| {\bf{u}} \right|}\nolimits^2 } \right\rangle$ is the equivalent isotropic displacement parameter, u describes the instantaneous displacement, and U _ij are the respective elements of the mean-square displacement tensor.

FIG. 4.

The crystal structure of Ca₂Mg₅Zn₅ as viewed (approximately) along the (a) [110] and (b) [001] directions. Zn atoms not belonging to the Sc₃Ni₁₁Si₄ structure type are marked by red arrows.

Powder XRD was carried out to test whether the selected grain for the single-crystal structure refinement was representative of the bulk sample. Approximately 5 g of the sample was ground into fine powder, and underwent XRD scanning from 10 to 90°. Peaks from Ca₂Mg₅Zn₅, matched by the lattice parameters determined by the single-crystal x-ray experiment, were found to dominate the diffraction pattern. Reflections from elemental Mg were also identified. Rietveld refinement with a starting model taken from the single-crystal investigation gave a reasonable fit (see Fig. 5) and essentially confirmed the single-crystal results. Furthermore, the lattice parameters and atomic coordinates of Ca₂Mg₅Zn₅ from single-crystal and powder refinements differed only by a few 0.01 Å, the site occupation factors agreed within 5%, and the total composition could be reproduced with deviations of less than 1% for each element. The Zn content in the Mg phase was found to be 3.4 at.%, which is also in good agreement with the EDX measurements mentioned above. However, it should be noted that the four shoulders (indicated by arrows in Fig. 5) could not be identified. This may be the result of unidentified phase(s), but considering that these shoulders were only observed near the strongest reflections, and that there were no other unidentified peaks, it is likely that they were caused by strain effects, particle size effects, oxidation of the powder, or a combination of these. The affected volume fraction is estimated to be a few percent. The R_wp value of the Rietveld full pattern fit was 11.3%. It is well-known that powder XRD refinements are less reliable than single-crystal investigations, and that this is particularly true in the presence of disorder and unidentified phases. Therefore we only report single-crystal refinement results here.

FIG. 5.

Results from the Rietveld refinement of the powder sample. Blue symbols: observed (experimental) intensities; green line: calculated intensities; turquoise line: I _obs − I _calc; red line: estimated background. Blue and red markers indicate reflection positions of the two refined phases. Arrows in the inset show some unidentified shoulders which may have resulted from variations in strain and particle size, or minor oxidation of the powder.

Because the structural information on Ca₂Mg₅Zn₅ was determined from scratch using single-crystal XRD and reported here for the first time, TEM SADP analysis was performed to match the experimental patterns obtained from the key alloy (Ca_12.40Mg_58.15Zn_29.45) with those simulated by jEMS, using the structural information from Table III. Visual comparisons between the experimental and simulated SADPs are presented in Fig. 6, for the zone axes (a) $\left[ {0001} \right]$ , (b) $\left[ {2\bar 1\bar 16} \right]$ , (c) $\left[ {10\bar 12} \right]$ , and (d) $\left[ {2\bar 1\bar 13} \right]$ . Note that images are presented in inverted contrast for better readability and that the size of the diffraction spots in the simulated patterns is proportional to their intensity. There appears to be an excellent match between the experimental and simulated patterns for all zone axes considered, confirming the identification of the new phase obtained from XRD. In Fig. 6(d), double diffraction from the Mg matrix and Ca₂Mg₅Zn₅ was considered, where the red diffraction spots in the simulated pattern resulted from the Mg matrix. When viewed together with the diffraction pattern of Ca₂Mg₅Zn₅, there was also good agreement between the experimental and simulated patterns thus confirming the derived structure for Ca₂Mg₅Zn₅.

FIG. 6.

SADPs, in inverted contrast, of Ca₂Mg₅Zn₅ crystals observed in the key alloy Ca_12.40Mg_58.15Zn_29.45. Four different crystallographic directions are shown: (a) $\left[ {0001} \right]$ , (b) $\left[ {2\bar 1\bar 16} \right]$ , (c) $\left[ {10\bar 12} \right]$ , and (d) $\left[ {2\bar 1\bar 13} \right]$ . The experimental SADPs are shown next to their corresponding simulated patterns using jEMS (displayed in blue frames).

The experimentally obtained ternary phase from equilibrium cooling, Ca₂Mg₅Zn₅, and the subsequent refinement of its structure may have a profound impact on our understanding of the Mg–Zn–Ca system, which has largely been based on thermodynamic calculations. For example, the computational thermodynamic model developed by Brubaker and Liu ^{Reference Brubaker and Liu33} abandoned Ca₂Mg₅Zn₅ and assumed Ca₂Mg₆Zn₃ to be the only ternary intermetallic phase in the system. Alternative CALPHAD models ^{Reference Mezbahul-Islam, Zhang, Shekhar and Medraj34,Reference Wasiur-Rahman and Medraj35} considered up to four different ternary phases. In light of the new experimental data reported here, further CALPHAD optimization studies are suggested to determine the potential impact of the Ca₂Mg₅Zn₅ ternary intermetallic phase on the calculated Mg–Zn–Ca ternary system. This may also have important consequences for the development of new high strength, highly ductile, and biodegradable MgZnCa alloys. ^{Reference Hofstetter, Becker, Martinelli, Weinberg, Mingler, Kilian, Pogatscher, Uggowitzer and Löffler6}