A large new family of filled skutterudites stabilized by electron count

Based on the interplay of theory and experiment, a large new family of filled group 9 (Co, Rh and Ir) skutterudites is designed and synthesized. The new materials fill the empty cages in the structures of the known binary CoSb3, RhSb3 and IrSb3 skutterudites with alkaline, alkaline earth, and rare earth atoms to create compounds of the type AyB4X12; A atoms fill the cages to a fraction y, B are the group 9 transition metals, and X is a mixture of electronegative main group elements chosen to achieve chemical stability by adjusting the electron counts to electron-precise values. Forty-three new compounds are reported, antimony-tin and phosphorous-silicon based, with 63 compositional variations presented. The new family of compounds is large and general. The results described here can be extended to the synthesis of hundreds of new group 9 filled skutterudites.

Because the electronic and magnetic properties of solids depend critically on the actual atoms making up the compounds in addition to the compound's electron count, expanding the filled skutterudites from primarily group 8 to group 9 metal based compounds, as we have done in this study, affords the opportunity to observe many new physical properties.
Here we report that an extremely large family of filled skutterudites based on the group 9 metals Co, Rh and Ir can be chemically stabilized if the electron count is stabilized by partial substitution on the X ion site to yield electron-precise formulas.
These new filled skutterudites were designed by filling or partial filling the cages in the binary CoSb 3 , RhSb 3 and IrSb 3 frameworks with alkali (Li, Na, K), alkaline earth metal (Ca, Sr, Ba), and rare earth atoms (La, Ce, Pr, Nd, Gd, Yb), by compensating for the extra positive charge of the filling ion by partial substitution of Sn on the Sb site or Si on the P site to yield electron-precise formulas, guided by the predictions of first principles electronic structure calculations. The pure antimonides and pure phosphides (with no Sb/Ns or P/Si mixing) are not stable, but we find that electron precise formulas are not strictly required for compound stability in some cases. In addition to their synthesis, crystal structures and electronic structures, we briefly survey the magnetic properties of selected members of this large new family of compounds.

Results and Discussion
The prototypical binary BX 3 skutterudites with B = Co, Rh, and Ir are nonmagnetic semiconductors, with a Zintl electron count of 24 valence electrons per formula unit. (The 24 electron count rule per BX 3 unit for semiconducting behavior also holds for a recently proposed non-Zintl view of the formal electron configuration of the constituent elements in skutterudites 52 .) Filled AB 4 X 12 skutterudites that have B in a d 6 low spin configuration and a p 6 X atom configuration are expected to be at electron precise, semiconducting, non-magnetic compounds when they have a valence electron count of 96 (4 x 24) per formula unit. The power of the Zintl concept is that these simple electron counting rules enable the prediction of new thermodynamically stable compounds. A Cobalt-based filled skutterudite compound can be stabilized, for example, by using Ba as an A 2+ ion and compensating for the added electrons by removing 2 electrons from the X ion site, and thus BaCo 4 Sb 10 Sn 2 should be stable, and an electron precise, non-magnetic semiconductor. In the same fashion the Rh and Ir compounds can be obtained as can the filled skutterudite semiconductors for any A ion.
This simple electron count principle is confirmed by our ab-initio calculations for several examples in this family, which are shown in Figure 2a. The valence band consists of strongly hybridized metal-d and Pnictogen-p states and is separated by a gap from the conduction band showing the expected semiconducting character.
Surprisingly, the electronic structure of the Rh variant calculated with the GGA is semimetallic, with a small valence band -conduction band overlap. Use of the MBJ functional in the calculation, which accounts for electronic correlations more effectively, on the other hand opens a band gap and predicts the Rh variant to be a semiconductor, although the predicted band gap is very small. This predicted anomalous behavior of the filled skutterudite variants based on Rh, which can be attributed to special characteristics of the Rh-Sb electronic interactions, would be of interest to verify through experiment.
The electron counting principle employed is similarly supported by the valence orbital region of the electronic DOS curves for the hypothetical compounds with formulas (simplified to allow for the calculations through the omission of Sn/Sb mixing) "LaCo/Rh/Ir 4 Sb 12 " (Figure 2b)  The compositions of the filled skutterudites was determined by standard solid state phase equilibria methods, which, using the characterization of reaction products by powder X-ray diffraction, especially for the heavy elements in compounds such as these, is quite sensitive to the presence of impurities. In addition, the crystal structures of all 63 variants reported here were determined by quantitative fits to the powder diffraction patterns.  Table 1 with the remainder given in the supplemental information in Table 1S. The cubic crystallographic cell parameters for all the compositions synthesized, determined by least squares fitting to 20 or more observed reflections in the powder XRD patterns using profile fits, are found in Table 2. for the lanthanides, has been used to describe the Ln site disorder for Os-based pyrochlores 55 . The very low thermal conductivity of the filled skutterudites based on group 8 elements has been attributed to the anharmonic thermal vibration of the A site ion "rattling" in the skutterudite cage 14,56 , implying a tendency toward off-center positioning of the A site ion in the cages in this family, in agreement with our structural refinements and those where the A site ion disorder is modelled by large thermal parameters. It is not yet known whether the A site disordered position model we observe here is a general feature of all larger filled skutterudites. Whether the disorder is static or dynamic or a combination of both, and whether there is any complex defect chemistry present in some of these compounds, would be of interest to study further by other structural characterization methods.
The important structural features of the filled skutterudites studied here are presented in Figure 4. This figure shows the geometry of the A site and group 9 site polyhedra and the manner in which they share faces (leading to the strong antibonding interactions described above), the displacements of the A site ions from the cage centers, and the X 4 squares. The lower part of the figure compares the X-X bond lengths within the slightly distorted X 4 antimonide and phosphide squares and the B-X bond lengths (all 6 are equivalent) for several representative members of the group 9 filled skutterudites. Neither the X 4 "squares" nor the BX 6 octahedra are constrained by symmetry to be perfect, and they are not, though they are all close and the distortions are relatively small. Because the same electron count is nearly perfectly maintained for all the compounds studied quantitatively here, large bond-length differences in these critical structural components are not expected to occur.
Filled skutterudites based on rare earth metals and group 8 transition elements show diverse and interesting rare earth magnetism, and they have been widely studied; see e.g 32,33,35 . In order to generally survey the magnetic properties for our new filled group 9 based skutterudites, the magnetizations of all the rare earth based filled skutterudites were measured from 2 to 200 K with an applied magnetic field of From these fits, the effective magnetic moment (P eff ) per Ln (Ln = La, Ce, Pr, Nd, Gd, and Yb) ion was obtained by using P eff = (8C) 1/2 . The thus derived basic magnetic characteristics (i.e.  cw and P eff ) for all our new compounds containing magnetic rare earths are summarized in Table 2. Fitting the magnetic susceptibility using the Curie Weiss law in the range 100 to 200 K, we obtain the effective moments for Pr 0.9 Co 4 Sb 10.2 Sn 1.8, Pr 0.9 Rh 4 Sb 10.2 Sn 1.8 , and Pr 0.9 Ir 4 Sb 10.2 Sn 1.8 of P eff = 3.43, 3.66 and 3.66  B /Pr, respectively, close to the effective moment value expected for the free Pr 3+ free ion (P eff = 3.58  B /Pr). In addition, the magnetic susceptibility data show a broad peak at around 3.5 K for Pr 0.9 Rh 4 Sb 10.2 Sn 1.8 (Figure 5d), which implies the onset of antiferromagnetic ordering. In order to better estimate the Neel temperature, we follow standard procedure 57-59 and plot d(χT)/dT (inset, Figure 5d). The maximum of d(χT)/dT is observed at 3.5 K, which can be taken as T N . Similar fits were performed for all the rare earth compounds synthesized. The new filled group 9 skutterudites that do not contain rare earths display temperature independent susceptibilities, with the exception of weak "Curie tails" at low temperatures due to the presence of impurity spins. Almost all of the intrinsic susceptibilities are in fact diamagnetic for these materials, indicating that they are dominated by core diamagnetism, as expected from the electron-precise formulas. The magnetic susceptibilities at 150 K for the non-magnetic samples, also presented in Table 2, are between -0.001 and -0.01 emu per mol formula unit. This indicates that the compounds do not display local moment behavior, but rather band behavior, even for the Co variants. More detailed study of the magnetic properties of the group 9 filled skutterudites will be of future interest.

Calculation, Synthesis and Experimental
The Experimental lattice constants were used and the free internal parameters were optimized by minimizing the forces. ternaries were first made by arc-melting the elements. Weight loss in this process was less than 1%. The as-prepared binaries or ternaries were then ground into powder and mixed with the appropriate stoichiometric quantities of elemental red P (99.9 %), pressed into pellets and then sealed in clean evacuated quartz ampoules. These ampoules were slowly heated to 400 o C, held for 10 h, and then slowly heated to 700 o C and held for 10 h; they were then slowly heated to 900 o C and held overnight.
Finally, they were thoroughly ground into a powder, repressed into pellets, and reheated at 900 o C and held there overnight.
The single phase compositions determined in the phase equilibria studies were specified by powder X-ray diffraction (XRD) using a Bruker D8 Focus diffractometer with Cu K radiation and a graphite diffracted beam monochromator. The initial structural model for the structures that were quantitatively refined from the powder diffraction data was taken from that of LaRu 4 Sb 12 69 . The FullProf software suite was used for the Rietveld refinements. Peak shapes were modeled with the Thompson-Cox-Hastings pseudo-Voight profile convoluted with axial divergence asymmetry.
The background was modeled with a Chebychev Polynomial. The structures were refined in space group Im with Co, Rh or Ir in the 8c sites and Sb/Sn in the 24g sites.
The position of the A site ion was dependent on the specific compound, as described.
In the final structural models, the structural parameters refined were the two positional coordinates (x and y) of the X atoms in position 24g, and when necessary,   58 Table 1. Structural parameters for refined crystal structures of selected filled skutterudites (and comparison to binary skutterudites). Table 2. Cubic cell parameters and selected physical properties of group 9-based filled skutterudites. Table 1S. Structural parameters for refined crystal structures of filled skutterudites (and comparison to binary skutterudites).
Figure1S. The Rietveld refinements of the laboratory powder XRD data for filled skutterudites (and comparison to binary skutterudites).    Figure1S. The Rietveld refinements of the laboratory powder XRD data for filled skutterudites (and comparison to binary skutterudites). All patterns were collected with Cu-k at 300 K.