The study of kagome lattice materials is an exciting topic in condensed matter due to their intrinsic geometric frustration, strong electron correlations, and electronic structure topology. 1,2 In particular, kagome materials have electronic band structure features that are symmetry protected, typically near or at the Fermi level. The kagome-derived bands are filled as flat bands (electrons with constant kinetic energy), Dirac cones (electrons avoiding a crossing point), and van Hove singularities (leading to enhanced electron-electron interactions). Among kagome materials, compounds with the formula LnM 6 X 6 (Ln = lanthanide; M = transition metal, X = groups 13 and 14) host a wide range of elemental combinations and set the stage for tuning of physical properties. As there are a diverse number of elements that can be substituted and there are two independent sublattices that can host a magnetic ion, kagome materials can host complex and noncollinear magnetic states. Several Mn 3 and Fe 4 analogues order magnetically with temperatures up to 400, 5 while Cr does not order. When magnetic lanthanides are present in these analogues, the two magnetic lattices (lanthanide and transition metal) couple, leading to interesting noncollinear or ferrimagnetic behavior, manifesting in interesting spin textures. 6 TbMn 6 Sn 6 exemplifies the exotic phenomena yielded from mixing topological electronic states and magnetism as it manifests a Chern gap. 7 The nonmagnetic Sc, 8,9 Y, 10 and Lu 11 LnMn 6 Sn 6 analogues host diverse magnetic states dominated by the Mn sublattice, such as magnetic spin chirality, long-range double-cone spin structure, and incom-mensurate antiferromagnetic arrangements. However, the V and Co analogues are the only families to exhibit no magnetic contribution from the transition metal. Nonmagnetic kagome metal analogues of LnV 6 Sn 6 have been identified as an ideal platform to study the interplay between lanthanide magnetism and kagome-derived band structures, such as noncollinear spin structures, 12 charge density waves, 13 and quantum critical behavior. 14 In contrast, the Co analogues have not been adequately examined: preliminary magnetic measurements of polycrystalline Co analogues have only been measured to 90 K with no indication of magnetic ordering. 15,16 Furthermore, it is important to consider that this structure may be susceptible to subtle variations (e.g., vacancies and superstructures), which would be expected to strongly affect the electronic and magnetic properties.
Recently, we discovered a new hybrid structure of the CoSn and YCo 6 Ge 6 structure types Yb 0.5 Co 3 Ge 3 . 17,18 Unusual magnetic anisotropy is observed at low temperatures for Yb 0.5 Co 3 Ge 3 , where the magnetic susceptibility for H//c exhibits a noticeable increase near 20 K, while a similar feature is not observed when measured along H//a and is proposed to be an indication of spin canting or reorientation. 17 For single crystals, the effective magnetic moment of μ eff = 4.23 μ B is less than the spin-only moment of Yb 3+ (4.54 μ B ). However, magnetization for H//c (the easy axis) shows a saturation moment of ∼1 μ B at 2 K. Recently, the presence of a kink at T ∼ 95 K in the electrical resistivity, coupled with a structural phase transition involving a change in space group from P6/mmm to P6 3 /m, was also found in Yb 0.5 Co 3 Ge 3 . 18 The combined structural phase transition and kink in the resistivity are characteristic of charge density waves in kagome metals, similar to what is seen for FeGe 19 and AV 3 Sb 5 . 20 The structural distortion in Yb 0.5 Co 3 Ge 3 is similar to the twisting of the equilateral triangles observed in MgCo 6 Ge 6 with the kagome nets maintaining planarity. 21 The distorted hexagons of the kagome net break two mirror planes while preserving inversion symmetry with a doubling of the unit cell along the caxis. These results suggest that the broader Ln 0.5 Co 3 Ge 3 family may provide a platform for studying the complex interplay between f-electron magnetism, electronic instabilities, and structural complexity. In order to further investigate this possibility, we present the crystal growth, single-crystal structural determination, magnetic properties, heat capacity, and resistivity on the single crystalline analogues Y x Co 3 (Ge 1-y Sn y ) 3 (x = 0.33, y = 0.11) and Gd x Co 3 (Ge 1-y Sn y ) 3 (x = 0.11, y = 0.133). Canfield crucible set, and sealed in a fused silica tube with ∼1/3 atm of Argon gas. After ramping the temperature up at 100 °C/h, the Y sample was dwelled at 1175 °C for 6 h and was cooled to 500 °C at a rate of 5 °C/h. The ampule with Gd was heated to 1000 °C at a rate of 100 °C/h and dwelled for 24 h. The system was then cooled to 815 °C at a rate of 2 °C/h. Both ampules were removed, inverted, and centrifuged at 3000 rpm. The residual Sn flux was etched from the crystals using 1:1 HCl/H 2 O. The synthesis of both analogues resulted in a high yield of rod-like morphology of the desired product. Single crystals up to 6 mm in length and a thickness of ∼0.4 mm were grown, with examples shown in Figure 1. A small amount of the LnCo 2 Ge 2 impurity can be visually identified and mechanically separated from the desired phase.
The crystal structures of the compounds were determined from single-crystal fragments using a Bruker D8 Quest Kappa single-crystal X-ray diffractometer equipped with an IμS microfocus source (Mo K α , λ = 0.71073 Å), a HELIOS optics monochromator, and a PHOTON III CPAD detector. The diffraction datasets were integrated using the Bruker SAINT program, and an absorption correction was applied to the intensities with a multiscan method in SADABS 2016/2. 22 The preliminary starting models were obtained using the intrinsic phasing method in SHELXT. 23 The indexed precession images from our datasets are available in the Supporting Information for both analogues (Figure S3).
For our preliminary structural models, Ge atoms were initially assigned to the Wyckoff sites 2c, 2e, and 1b (atomic coordinates are provided in Table 1) similar to the model previously reported in Yb 0.5 Co 3 Ge 3 ; 17 however, the 1b site had an elevated amount of residual electron density in comparison to the other residual two Ge sites. Sn was assigned at the 1b site, effectively reducing refinement statistics and residual electron density. Unlike our previously reported Yb 0.5 Co 3 Ge 3 model with less than 8% occupancy of Ge on the 1b site, Y x Co 3 (Ge 1-y Sn y ) 3 (x = 0.33, y = 0.11) and Gd x Co 3 (Ge 1-y Sn y ) 3 (x = 0.3, y = 0.133) have Sn occupancies of ∼33 and ∼40%, respectively. The occupancies of the Ln and Ge2 sites were constrained to be equivalent after the independent refinements of the Y analogue yielded statistically similar occupancies of 0.332(3) for Y and of 0.325 (5) for Ge2, and the Gd analogue yielded occupancies of 0.2905 (16) for Gd and of 0.292(5) for Ge2. Additionally, an EADP command was initially used on the Ge2 and Sn2 sites, where both residual electron density of 1.98 e Å -3 and a hole of -2.69 e Å -3 were present <0.50 Å away from Ge2. However, removing the command resulted in a peak of 0.56 e Å -3 and a hole of -0.782 e Å -3 . While these modeling differences are minor, the removal of the constraint yields better figures of merit and a better fit with the EDS values collected. To corroborate this observation, observed Fourier maps of the disordered regions were calculated, and there is a prolate ellipsoid for the Sn2 site and not for the Ge2 site, shown in Figure S5.
Crystallographic data and refinement parameters are reported in Tables S1 andS2. Data were also collected at 90 K; crystallographic data, refinement parameters, and atomic positions are provided in Tables S3 andS4.
Single crystals of Y 0.33 Co 3 Ge 2.67 Sn 0.33 and Gd 0.30 Co 3 Ge 2.60 Sn 0.40 were analyzed using energy-dispersive (X-ray) spectroscopy (EDS) on a VERSA 3D focused ion beam scanning electron microscope with an acceleration voltage of 20 kV and a spot size of 7 to confirm the presence of Sn in the sample. Data were collected for 5 spots on Y x Co 3 (Ge 1-y Sn y ) 3 , and three area scans (13.23 × 2.86 μm) were used for Gd x Co 3 (Ge 1-y Sn y ) 3 . The weight percentages obtained from EDS resulted in the atomic formulas Y 0.37(1) Co 3.00(3) Ge 2.61(4) Sn 0.39(1) and Gd 0.31(2) Co 3.00(2) Ge 2.52(2) Sn 0.24 (1) , which are in good agreement with composition from single-crystal refinements Y 0.33 Co 3 Ge 2.67 Sn 0.33 and Gd 0.30 Co 3 Ge 2.60 Sn 0.40 , respectively. Values were normalized to Co as it does not participate in the disorder within the system. In addition, the powder diffraction data for ground single crystals are provided in Figures S2 andS3. For simplicity, Y 0.33 Co 3 Ge 2.67 Sn 0.33 and Gd 0.30 Co 3 Ge 2.60 Sn 0.40 will be denoted as Y x Co 3 (Ge 1-y Sn y ) 3 and Gd x Co 3 (Ge 1-y Sn y ) 3 , respectively.
2.3. Property Measurements. Temperature-dependent magnetization measurements were carried out for T = 1.8-300 K under magnetic fields of H = 0.1 T applied parallel (//) and perpendicular (⊥) to the crystallographic c-axis using a Quantum Design VSM Magnetic Property Measurement System. The specimens were attached to the quartz rods using GE-varnish. The isothermal magnetization measurements were also performed for H < 7 T. The specific heat (C) measurements were performed for T = 1.8-200 K in a Quantum Design Physical Properties Measurement System using a conventional thermal relaxation technique. Electrical resistivity (ρ) measurements for Y x Co 3 (Ge 1-y Sn y ) 3 (x = 0.33, y = 0.11) over the temperature range of T = 1.8-300 K were performed in a four-wire configuration, where the platinum wires were spot-welded to the crystals in a standard four-wire configuration. The temperature dependence of resistivity of Gd x Co 3 (Ge 1-y Sn y ) 3 (x = 0.3, y = 0.133) was measured in an Oxford wet cryostat. The four-probe method with an AC current applied by a Zurich lock-in amplifier (MFLI) was used to measure the resistivity. 2). The framework of alternating Co kagome and Ge honeycomb nets along the c-axis is ordered; however, the channels formed through the alignment of hexagonal voids host disordered "stuffing atoms". The stuffing atoms consist of Gd (planar to the honeycomb layer), Sn2 (planar to the kagome layer), and Ge2 (out of plane with the kagome layer by ∼0.20 Å above or below). The atoms within the channel can be best broken down into three disordered components. These disordered components are a pair of Gd-Ge2 dimers (each pair ∼30% occupied) and a lone ∼40% occupied Sn2 as shown in Figure 3. The occupancies of these components are constrained to a total occupancy of 100%. The Ln x Co 3 (Ge 1-y Sn y ) 3 structure is considered to be a hybrid of the CoSn and YCo 6 Ge 6 structure types. The structure is reminiscent of CoSn when the Sn2 site is present and YCo 6 Ge 6 when either of the two Gd-Ge2 dimers are present. The hybrid of the two structure types leads to the presence of disorder.
Of the LnM 6 X 6 structure type, there are two major "stuffed" CoSn variants: the ordered HfFe 6 Ge 6 [a = 5.069(1) Å, c = 8.041(1) Å] 24 and the disordered YCo 6 Ge 6 [a = 5.074 Å and c = 3.908 Å] 25 structure types. Figure 4 shows the crystal structures of CoSn, 26 HfFe 6 Ge 6 , and YCo 6 Ge 6 to provide direct comparison between related phases.
CoSn consists of alternating Sn honeycomb nets and Sncentered Co kagome nets stacking along the c-axis. The alternating stacking nets form voids inside of a 20-atom polyhedron, as shown in Figure 5. The CoSn structure type, (Co 3 Sn)(Sn 2 ), can be divided into two alternating nets: the Co 3 Sn kagome nets and the Sn 2 honeycomb nets. A Ge-doping study of the CoSn structure reveals that the site preference doping can lead to a doubling of the cell. The CoSn structure type is preserved when Ge is in the majority or Ge:Sn ratio of ∼2:1. The honeycomb nets only contain Ge. Sn is mixed with Ge in the plane of the Co kagome nets, which can be expressed as (Co 3 Sn 1-y Ge y )(Ge 2 ). Conversely, when Sn is in the majority or a Sn:Ge ratio of ∼2:1, the c-axis of the CoSn structure type is doubled similar to what is observed in the HfFe 6 Ge 6 structure type. Sn is mixed with Ge in the honeycomb nets, while only Sn is in the plane of the Co kagome nets and is classified as (Co 3 Sn)(Sn 2-x Ge x ). This implies that we can consider the site preference of Ge doping as a preference for kagome or honeycomb net structures. For our new hybrid analogues (Ln = Y and Gd), a rare earth site in the Ge 2 honeycomb nets is being filled up to a maximum of half occupied, which is observed in the YCo 6 Ge 6 structure type. As the lanthanide site is being filled, the Sn site in the Co 3 Sn kagome nets is displaced above and below the plane of the kagome nets. In our case, the displaced positions are "preferably" occupied by Ge instead of Sn.
While the CoSn structure type is widely viewed as the stuffed host for the LnM 6 X 6 family, the CoGe binary is not isostructural. The ambient stable CoGe binary is a monoclinic distortion that breaks the kagome and honeycomb layers; however, it retains some of the motifs. The Sn-doping studies of the CoGe structure reveal the Sn site preference of the stuffing atom and stabilize the CoSn structure type. 27 Although it is unclear if the Sn doping in Y x Co 3 (Ge 1-y Sn y ) 3 (x = 0.33, y = 0.11) and Gd x Co 3 (Ge 1-y Sn y ) 3 (x = 0.3, y = 0.133) is a result of stability, it is interesting to note that the site preference is similar to that of its binary counterpart. Additionally, with the Sn substitution, we see that there is a stronger contribution of the CoSn structure type in the disorder compared to the undoped analogues. Y and Gd have Sn occupying the 1b in the CoSn component of the hybrid structure 33 and 30%, respectively, whereas the Yb analogue without Sn has a nearly negligible amount of Ge on the 1b site (∼8%). Seemingly, the incorporation of Sn in Ln x Co 3 (Ge 1-y Sn y ) 3 could be used as a tuning parameter to influence the hybrid structure's dominant component and possibly influence the amount of lanthanide stuffing. Although there is very little work done on Yb 0.5 Co 3 (Ge 1 Sn y ) 3 structure types, we look to doping studies within the HfFe 6 Ge 6 structure type to infer that the same type of doping is seen in ternary systems. Across multiple systems, there is a consistent site preference in small amounts of doping, where Sn prefers to dope within the stuffing atoms and Ge prefers the honeycomb net�consistent with our observation of doping. [28][29][30]
The temperature-dependent magnetic susceptibility data for the nonmagnetic rare earth analogue Y x Co 3 (Ge 1-y Sn y ) 3 (x = 0.33, y = 0.11) were collected on single crystals. These data reveal that Y x Co 3 (Ge 1-y Sn y ) 3 (x = 0.33, y = 0.11) is a temperature-independent paramagnet and Co does not carry a magnetic moment (Figure S4). Figure 6a shows anisotropic magnetic susceptibility and inverse magnetic susceptibility data of Gd x Co 3 (Ge 1-y Sn y ) 3 (x = 0.3, y = 0.133) , where the a-and c-axes have slightly different high-temperature behaviors. Unlike the previously reported data collected on polycrystalline GdCo 6 Ge 6 , 31 there is no magnetic ordering down to 1.8 K for a single crystal of Gd x Co 3 (Ge 1-y Sn y ) 3 (x = 0.3, y = 0.133). A Curie-Weiss fit to the data for temperatures T = 50-300 K yields an effective magnetic moment of μ eff = 7.84(1) μ B /Gd along the H//c direction and is consistent with the value expected for Gd 3+ moment (μ eff = 7.94 μ B /Gd). However, an elevated μ eff of 8.48 μ B is seen for measurements along the H//a direction. The Weiss constants for H//a and H//c are -4.83 and 0.08 K, respectively, indicating antiferromagnetic interactions along the a-axis and little magnetic frustration as θ/T N ∼ 3. 32 The low ordering temperature in the Gd analogue likely results from a weak magnetic exchange interaction between the Gd ions. The a i s o t r o p i c fi e l d -d e p e n d e n t m a g n e t i z a t i o n o f Gd x Co 3 (Ge 1-y Sn y ) 3 (x = 0.3, y = 0.133), measured at T = 1.8 K, is shown in Figure 6b, where both curves rise rapidly and begin to saturate in the vicinity of 2 T. The saturation moment for H//a is approximately 6.6 μ B , which is slightly less than a Gd 3+ J = 7/2 system (M sat = 7 μ B /Gd). Along the c-axis, there is an elevated magnetic saturation moment of 7.9 μ B , again implying that there is a magnetic contribution from the Co-kagome net: e.g., an elevated magnetic moment along the a-axis in comparison to the c-axis could indicate that the Comagnetism occurs planar to the kagome net. While the origin of the magnetism in the Co sublattice is unknown, Co magnetism is not seen for other analogues 15,16,31 and CoSn (and Ge-doped analogues), 27,33 and we speculate that the presence may be due to the impact of the Gd f-state on the electronic band structure. Further work (e.g., electronic structure calculations or a more systematic study of other Ln variants) is needed to answer this question.
Figure 7a shows the temperature-dependent heat capacity of Y x Co 3 (Ge 1-y Sn y ) 3 (x = 0.33, y = 0.11) and Gd x Co 3 (Ge 1-y Sn y ) 3 (x = 0.3, y = 0.133) measured at H = 0 T with a temperature range of 0.5-200 K. Both the Sommerfeld coefficients and the Debye temperatures were determined by fitting the data using 7c) yielded γ = 9 mJ/mol K 2 and β = 0.26 mJ/mol K 4 , consistent with the results from the integral Debye fit. In the heat capacity of Y x Co 3 (Ge 1-y Sn y ) 3 , there also is a weak kink at T ∼ 118 K, possibly representing a phase transition that is reminiscent of what is seen in MgCo 6 Ge 6 and Yb 0.5 Co 3 Ge 3 . 18,21 For those compounds, kinks are also seen in the resistivity and have been shown to be related to a structural phase transition. However, for Y x Co 3 (Ge 1-y Sn y ) 3 (x = 0.33, y = 0.11), low-temperature singlecrystal X-ray diffraction data were collected down to 90 K and provided no indication of a phase transition (Tables S3 andS4). In the Gd analogue, the low-temperature heat capacity measurements reveal a broad peak at T = 1.7 K, which indicates magnetic ordering. At higher temperatures, the heat capacities of the Y and Gd analogues have different behaviors, possibly indicating different phonon interactions or the presence of an itinerant cobalt magnetic contribution that was not observed in the Y analogue. Alternatively, changes in the disorder occupancy could lead to the change of phonon modes; however, the measurement of other lanthanide analogues is required to establish a trend. Figure 7b shows the electrical resistivity as a function of temperature. Also noteworthy is that ρ(T) for single crystalline of Y x Co 3 (Ge 1-y Sn y ) 3 (x = 0.33, y = 0.11) is lower by an order of magnitude than previously reported polycrystalline samples that were produced by arc melting. 15 Conventional metallic behavior is observed over the entire temperature range for both compounds. Interestingly, there is a broad hump in ρ(T) apexing near T = 120 K for the Y compound that roughly coincides with the feature that is seen in C/T. While this does not definitively establish the presence of a phase transition, the positive curvature is consistent with weak and gradual gapping of the Fermi energy. Finally, the sudden decrease of resistivity at ∼3 K is due to Sn inclusions; Sn contains a superconducting transition temperature of 3.7 K. 34
Herein, we are reporting the new hybrid structures Y x Co 3 (Ge 1-y Sn y ) 3 (x = 0.33, y = 0.11) and Gd x Co 3 (Ge 1-y Sn y ) 3 (x = 0.3, y = 0.133) with Sn incorporation that could ignite the investigation of Sn analogues and doping studies and invite the reinvestigation of other lanthanide analogues. The growth of single crystals allowed novel anisotropic magnetic measurements that reveal exclusive Co magnetic contribution along H//a in the Gd analogue, which is in-plane with the kagome net. The magnetic properties for the nonmagnetic Y analogue do not contain Co contributions, which we speculate could be due to the impact of Gd f-electrons. In addition to magnetic measurements, the heat capacity study also revealed a lowtemperature magnetic transition at ∼1.7 K in the Gd analogue; this transition, coupled with the insights of electronic kagome features, could result in new exotic states, justifying further theoretical and experimental investigations.
The ability to tune magnetism in the Ln x Co 3 (Ge 1-y Sn y ) 3 family with a twist distortion, [35][36][37] similar to the superconducting LaRu 3 Si 2 family, makes the nonmagnetic analogue a great platform to study the interplay of magnetism correlation. The kagome net means that magnetism has to deal with the electronic correlation in the system. Furthermore, this work illustrates the family of Ln x Co 3 (Ge 1-y Sn y ) 3 as a special platform to further investigate the interplay of magnetism and correlation and invites the discovery of structures of the Ln x Co 3 (Ge 1-y Sn y ) 3 type, where related structural complexity and magnetism are best elucidated via growth of large single crystals. We expect that new materials will advance the understanding of the interplay of electronic correlation, topology, and magnetism in these kagome compounds. The observation of magnetic ordering in an inherent topological kagome lattice would warrant further investigation of its effects on the band structure through theoretical and experimental methods.
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.inorgchem.3c02172.
Crystallographic data and refinement parameters of Y x Co 3 (Ge 1-y Sn y ) 3 and Gd x Co 3 (Ge 1-y Sn y ) 3 ; selected interatomic distances of Y x Co 3 (Ge 1-y Sn y ) 3 and Gd x Co 3 (Ge 1-y Sn y ) 3 ; low-temperature crystallographic
https://doi.org/10.1021/acs.inorgchem.3c02172Inorg. Chem. 2023, 62, 18049-18055