Ternary aurides RE₄Mg₃Au₁₀ (RE=Y, Nd, Sm, Gd–Dy) and their silver analogues

2.1 Synthesis

Starting materials for the syntheses of the RE₄Mg₃Ag₁₀ and RE₄Mg₃Au₁₀ phases were pieces of sublimed rare earth elements (Smart Elements, > 99.9%), silver and gold granules (Allgussa AG, > 99.9%), and a magnesium rod (Johnson Matthey, > 99.5%). The surface layer of the magnesium rod was removed on a turning lath in order to avoid oxidic impurities. The moisture-sensitive neodymium and samarium ingots were kept under dry argon in Schlenk tubes. Smaller pieces were cut under paraffin oil and washed with cyclohexane. Both the paraffin oil and cyclohexane were dried over sodium wire. The argon was purified over titanium sponge (900 K), silica gel, and molecular sieves.

The RE₄Mg₃Ag₁₀ and RE₄Mg₃Au₁₀ samples were prepared with the ideal elemental mixtures of 4RE: 3Mg: 10Ag(Au). The elements were arc-welded [13] in small niobium ampoules under an argon pressure of ca. 700 mbar. The ampoules were then placed in a water-cooled sample chamber of an induction furnace [14] (Typ TIG 2.5/300; Hüttinger Elektronik, Freiburg, Germany) and rapidly heated to ca. 1500 K. After 1 min, the temperature was slowly reduced to 1300 K, kept for 30 min, followed by cooling to 900 K within another 30 min. The tubes were kept at 900 K for two hours before quenching the annealing sequence by turning off the high-frequency generator.

The reaction products were mechanically separated from the niobium ampoules. No attack of the crucible material was visible. There were strong hints that the reactions precede peritectically. Quenching of the samples showed formation of considerable amounts of the binary phases REAg₂ and REAu₂. This tendency is even more pronounced in going to the heavier rare earth elements. In order to achieve better crystallinity of the samples and to reduce the amount of the binary phases, the polycrystalline products were sealed in evacuated silica ampoules and annealed at 823 K for 1 week. The polycrystalline RE₄Mg₃Ag₁₀ and RE₄Mg₃Au₁₀ samples are dark gray. Small single crystals appear light gray with metallic luster. The samples are stable in air over several weeks.

The three gadolinium-containing crystals originated from different samples. Phase analytical studies by metallography in combination with EDX first revealed a new phase with an approximate composition of 21 at% Gd: 14.5 at% Mg: 64.5 at% Au. A sample with that starting composition was first annealed at 1300 K for 1 min, then slowly cooled to 770 K within 25 min and kept at that temperature for another 2 h. The crystal taken from this sample had the composition Gd_5.61Mg_1.39Au₁₀. A sample of the starting composition 5 Gd:2 Mg:10 Au then yielded the crystal Gd_5.50Mg_1.50Au₁₀, and the ideal starting composition 4 Gd:3 Mg:10 Au (with a final annealing step of 7 d at 870 K) yielded Gd_4.43Mg_2.57Au₁₀.

2.2 EDX data

Semiquantitative EDX analyses of the single crystals studied on the diffractometers were carried out in variable pressure mode with a Zeiss EVO^® MA10 scanning electron microscope with gadolinium trifluoride, magnesium oxide, and gold as standards. No impurity elements were detected.

2.3 X-ray diffraction

The polycrystalline RE₄Mg₃Ag₁₀ and RE₄Mg₃Au₁₀ samples were studied by powder X-ray diffraction using the Guinier technique: imaging plate detector, Fujifilm BAS-1800, CuK_α1 radiation and α-quartz (a=491.30, c=540.46 pm) as an internal standard. The orthorhombic lattice parameters (Table 1) were refined from the Guinier powder data by least-squares fits. Correct indexing was ensured by comparison of the experimental patterns with calculated ones [15].

Small single-crystalline fragments were selected from the different gadolinium-based samples. The crystals were glued to small quartz fibers using beeswax and were characterized on a Buerger camera (using white Mo radiation) to check their quality. Intensity data of three different crystals were collected at room temperature on a Stoe IPDS-II image plate system (graphite-monochromatized Mo radiation; λ=71.073 pm) in oscillation mode. Numerical absorption corrections were applied to the data sets. Details about the data collections and the crystallographic parameters are summarized in Table 2.

2.4 Structure refinements

The three diffractometer data sets showed C-centered orthorhombic lattices, and the systematic extinctions were compatible with space group Cmca. The starting atomic positions were deduced from interpretations based on Direct Methods with shelxs-97 [16, 17], and the structures were refined using shelxl-97 [18, 19] (full-matrix least squares on F²) or jana2006 [20] with anisotropic atomic displacement parameters for all atoms. Inspection of the Pearson data base [21] for the Pearson code oC68 and the Wyckoff sequence g²f²eda readily revealed isotypism with the indium phase Ca₄In₃Au₁₀ [22]. In the following cycles, we then refined the three structures with the setting of the indium compound. Since the isotypic cadmium compound Ca₅Cd₂Au₁₀ [23] revealed Ca/Cd mixing on two of the cation sites, we carefully checked the occupancy parameters of the gadolinium sites. For the 4a and 8e sites, we observed pronounced Mg/Gd mixed occupancies for all three crystals. These occupancy parameters were refined as least-squares variables in the final cycles, leading to the compositions listed in Table 2. The final difference Fourier syntheses revealed no residual peaks. The refined atomic positions, displacement parameters, and interatomic distances (exemplarily for Gd_4.43Mg_2.57Au₁₀ with almost ideal composition) are given in Tables 3 and 4.

Further details of the crystal structure investigations may be obtained from Fachinformationszentrum Karlsruhe, 76344 Eggenstein-Leopoldshafen, Germany (fax: +49-7247-808-666; e-mail: crysdata@fiz-karlsruhe.de, http://www.fiz-karlsruhe.de/request_for_deposited_data.html) on quoting the deposition number CSD-428603 (Gd_4.43Mg_2.57Au₁₀), CSD-428601 (Gd_5.50Mg_1.50Au₁₀) and CSD-428602 (Gd_5.61Mg_1.39Au₁₀).

3.1 Crystal chemistry

The ternary phases RE₄Ag₁₀Mg₃ (RE=Y, Nd, Sm, Gd–Dy) and RE₄Au₁₀Mg₃ (RE=Y, Nd, Sm, Gd–Dy) crystallize with a ternary ordered version of the Zr₇Ni₁₀ type [24], similar to Ca₄In₃Au₁₀ [22] and Ca₅Cd₂Au₁₀ [23]. The cell volumes of the bulk phases decrease from the neodymium to the dysprosium compound (Table 1) as expected from the lanthanoid contraction. Both yttrium compounds fit within this series between the terbium and dysprosium representatives, similar to many other series of rare earth-based magnesium intermetallics [1]. The gold compounds have slightly smaller cell volumes than the silver compounds, a consequence of the relativistic contraction. Along with Ca₄In₃Au₁₀ [22] and Ca₅Cd₂Au₁₀ [23], the magnesium compounds are the only ternary representatives of the Zr₇Ni₁₀ type. Besides the prototype [24–26], Hf₇Ni₁₀ [27], Zr₇Cu₁₀ and Hf₇Cu₁₀ [28], the solid solution Zr₇Ni_10–xCu_x [29], Hf₇Au₁₀ [30], and Zr₇Pt₁₀ [31] have been reported.

Of all ternaries, complete ordering only occurs for Ca₄In₃Au₁₀ [22]. The calcium and indium atoms are ordered on the Zr3 and Zr1 (Ca1 (8d) and Ca2 (8f)), respectively, Zr4 and Zr2 (In1 (4a) and In2 (8e)) sites. While the 8d and 8f sites are fully occupied by gadolinium in the presently studied crystals, the sites 4a and 8e show substantial mixed occupancies with magnesium. Thus, one can write a general formula Gd_4+xMg_3–xAu₁₀ with x up to 1.50. Such rare earth-magnesium mixed occupancies also occur in many of the Gd₄RhIn-type [32] phases RE₄TMg [2, 33–35]. In view of the differences in chemical bonding between the binary prototype Zr₇Ni₁₀ and the ternaries described herein, one should call their relationship rather isopointal [36, 37] than isotypic.

The crystal chemical details and the chemical bonding have been described in detail for Ca₄In₃Au₁₀ [22] and Ca₅Cd₂Au₁₀ [23]. In the following discussion, we only refer to the single-crystal data of Gd_4.43Mg_2.57Au₁₀, which has almost the ideal composition. For reasons of simplicity, we first neglect the Mg/Gd mixing in the crystal chemical description. Then we discuss trends within the solid solution Gd_4+xMg_3–xAu₁₀. The unit cell of Gd_4.43Mg_2.57Au₁₀ is presented in Fig. 1. Due to the high gold content, one observes pronounced Au–Au interactions within the structure. The Au–Au distances between the three crystallographically independent gold atoms range from 278 to 297 pm, similar to fcc gold (288 pm) [38]. This leads to a pronounced two-dimensional gold substructure in the bc direction. These layers are stacked along the a axis. The shortest Au–Au distance between these blocks is at 309 pm and can be considered as a secondary interaction. This trend in Au–Au distances is similar to that for Ca₄In₃Au₁₀ [22] and Ca₅Cd₂Au₁₀ [23].

Fig. 1

View of the Gd_4.43Mg_2.57Au₁₀ structure approximately along the (011) direction. Gadolinium, gold and magnesium atoms are drawn as medium gray, blue, and magenta circles, respectively. The three-dimensional gold network is emphasized and the atom designations are given at the lower right-hand part. Note that the Mg1 and Mg2 sites show mixing with gadolinium.

The polyanionic gold blocks leave cavities that are filled by the Mg1 atoms. Adjacent blocks are connected by the Mg2 and Gd1 atoms (Fig. 1). The Mg1 atoms have eight nearest gold neighbors at Mg1–Au distances in the range of 283–285 pm, slightly longer than the sum of the covalent radii for gold and magnesium of 270 pm [39]. This is similar to the equiatomic compounds EuAuMg (TiNiSi type, 284–294 pm Au–Mg) [40] and GdAuMg (ZrNiAl type, 278–291 pm Au–Mg) [41]. We can thus assume medium Au–Mg bond strength. This is different from Ca₄In₃Au₁₀ [22], where the highest overlap populations were observed for the Au–In interactions. This difference follows the general trends for isotypic magnesium and indium compounds, which always show weaker T–Mg bonding as compared with T–In bonding [42]. The Gd2 atoms within the gold blocks have the higher coordination number (CN) 17, a consequence of their larger size. Again, they have 11 gold atoms as nearest neighbors.

While the binary rare earth gold compounds are sensitive to moisture, the ternary compounds reported herein are air stable. Introducing magnesium within the 3D [Mg₃Au₁₀] network leads to substantial covalent Au–Mg bonding, thus stabilizing our compounds. This is similar to many other ternary gold-based stannides, indides, and cadmium compounds.

An alternative description of the Gd_4.43Mg_2.57Au₁₀ structure is possible by assuming a dense packing of two polyhedra (Fig. 2), which can be considered as basic building units. The first polyhedron is built around the Mg1 atoms. Eight gold and four gadolinium atoms coordinate to Mg1 in the form of a strongly flattened cuboctahedron. These CN12 polyhedra are condensed via common corners in bc direction at x=0 and x=1/2, respectively. Adjacent layers of these polyhedra are located in the gaps, a consequence of the C-centered lattice. The Gd1@Au₁₀Mg₄ polyhedra are the second type of basic building units. Always two Gd1@Au₁₀Mg₄ polyhedra are condensed to the Mg1@Au₈Gd₄ polyhedra via common rectangular faces. Adjacent Gd1@Au₁₀Mg₄ polyhedra are connected via a common edge of gold atoms. These double units are further condensed with Mg1@Au₈Gd₄ polyhedra of the next layer via gold atoms.

Fig. 2

The structure of Gd_4.43Mg_2.57Au₁₀. Gadolinium, gold and magnesium atoms are drawn as medium gray, blue, and magenta circles, respectively. The Mg1@Au₈Gd₄ (light gray) and Gd1@Au₁₀Mg₄ (medium gray) polyhedra are emphasized. Note that the Mg1 and Mg2 sites show mixing with gadolinium. For better visibility only two Gd1@Au₁₀Mg₄ polyhedra are drawn.

Now we turn back to the Mg/Gd mixing. Principally, at first sight, ordering of the four zirconium sites of the Zr₇Ni₁₀ type would allow mixing for all sites. However, the four sites have different coordination numbers. For Gd1, Gd2, Mg1, and Mg2 of Gd_4.43Mg_2.57Au₁₀, we observe CN16, CN17, CN12, and CN14. Thus, from a geometrical point of view (the magnesium atoms are smaller than the gadolinium atoms), it is understandable that Mg/Gd mixing is only observed for the two sites with the smaller coordination number.

The second point concerns the Bader charges observed for the indide auride Ca₄In₃Au₁₀ [22]: +1.49 and +1.39 for Ca1 and Ca2, –0.18 and –0.34 for In1 and In2, and –0.46, –0.47, and –0.62 for Au1, Au2, and Au3, respectively. In the compounds presented herein, the magnesium atoms take the indide sites and the gadolinium atoms fill the calcium positions. In contrast to the indium atoms, the magnesium atoms will surely have a weak positive charge, already manifesting the bonding differences between Ca₄In₃Au₁₀ and our compounds.

The degree of Mg/Gd mixing strongly depends on the synthesis conditions. The crystals studied herein were taken from three samples with different starting compositions and annealing sequences. Quenching of the samples from high temperature gives only low yields of the 4-3-10 samples besides formation of the binary REAu₂ and REAg₂, and yet unknown phases. Interestingly, the crystals taken from these samples showed the highest gadolinium contents, while the reaction yields increased with longer annealing sequences at lower temperatures along with higher magnesium occupancy on the CN12 and CN 14 sites.

Our phase analytical studies gave hints also for 4-3-10 with the smaller and larger rare earth elements; however, no samples have yet been obtained with high yields. Further synthetic studies are in progress in order to exemplarily study one of the silver- and gold-containing phases with respect to the existence range of the solid solutions RE_4+xMg_3–xAg₁₀ and RE_4+xMg_3–xAu₁₀, as well as the influence of the annealing temperature on the phase formation.

Finally, we briefly comment on the isotypic silver compounds. So far, the RE-Ag-Mg systems have only scarcely been studied [43–51]. An isothermal section is only reported for the lanthanum-based system [50]. Most studies were performed in order to determine the structures and the physical properties of the equiatomic compounds REAgMg [44–48]. The RE₄Mg₃Ag₁₀ phases reported herein are isotypic with the recently reported compound La₄Mg₃Ag₁₀ [49]. The latter also shows a small degree of Mg/La mixing, and the results of the bonding analysis are comparable to that of Ca₄In₃Au₁₀ [22].