The Weak 3D Topological Insulator Bi12Rh3Sn3I9

Abstract Topological insulators (TIs) gained high interest due to their protected electronic surface states that allow dissipation‐free electron and information transport. In consequence, TIs are recommended as materials for spintronics and quantum computing. Yet, the number of well‐characterized TIs is rather limited. To contribute to this field of research, we focused on new bismuth‐based subiodides and recently succeeded in synthesizing a new compound Bi12Rh3Sn3I9, which is structurally closely related to Bi14Rh3I9 – a stable, layered material. In fact, Bi14Rh3I9 is the first experimentally supported weak 3D TI. Both structures are composed of well‐defined intermetallic layers of ∞ 2[(Bi4Rh)3I]2+ with topologically protected electronic edge‐states. The fundamental difference between Bi14Rh3I9 and Bi12Rh3Sn3I9 lies in the composition and the arrangement of the anionic spacer. While the intermetallic 2D TI layers in Bi14Rh3I9 are isolated by ∞ 1[Bi2I8]2− chains, the isoelectronic substitution of bismuth(III) with tin(II) leads to ∞ 2[Sn3I8]2− layers as anionic spacers. First transport experiments support the 2D character of this material class and revealed metallic conductivity.


Introduction
Large efforts have been devoted to the synthesis and to the physicalf undamentals of topological insulators( TIs) to investi-gate their intriguing physical properties. [1,2] While being bulk semiconductors, TIs can host protected metallic states on their surfaces. Electrons in these topologically protected surface states are sheltered against scattering. Their spin and momentum are locked orthogonally to their propagation direction. [3,4] This effect is manifested in almost dissipation-free electron transport. Thus, TIs are envisioned as promising candidatesi n the field of electronics forh igh-performance spin field-effect transistors. [5] and quantum bits in quantum computing. [6] The discoveries started with the investigationso nt he quantum Hall and the fractional quantum Hall effect, years ago. Both impressivep henomenaw ere awarded with the Nobel Prize in Physics in 1985 and 1998, respectively,a nd led to comprehensive research in the following years. In 2006 and2 007, the theoreticalp rediction and the experimental observation of the quantum spin Hall effect in HgTew ells marked the beginning of the investigation on TIs. [7,8] The focus was put on the electronic state, which is similar to the quantum Hall effect, but emerged due to the inherents pin-orbit coupling insteado f the external magnetic field. The existence of aq uantum effect at room temperature reveals ag reat potential for applications and stimulated aw ide search for new TIs.
In the recent past, our group has pursued research in TIs and has contributedt ot he synthesis and characterization of, for example, Bi 14 Rh 3 I 9 , [9][10][11][12][13] Bi n Te I( n = 1, 2), [14,15] and MnBi 2 Te 4 . [16] Amongt hose, Bi 14 Rh 3 I 9 is known as the first experimentally verifiedw eak 3D TI. Its large topologically non-trivial band gap of 210 meV is generated by strong spin-orbit coupling (SOC) and assurest hat the quantum state persists at room temperature and up to the decomposition temperature of the compound. Topological edge states have been observed by scanning tunnelling microscopyo ns urface stepso fB i 14 Rh 3 I 9 . [10] [a] M. LÞ  Bi 14 Rh 3 I 9 comprises intermetallic kagome networks of rhodiumcentred bismuth cubes that sharec ommone dges forming hexagonal-prismatic gaps, in which iodine atoms are located ( Figure 1). [9,17] As previous works have demonstrated, the main covalentbonding system is located inside the 1 2 [(Bi 4 Rh) 3 I] 2 + intermetallic layer.Each of these layers was characterized as a2 D TI, if appropriate charge compensation is applied. [12] These layers together with the topologically trivial anionic spacer of 1 1 [Bi 2 I 8 ] 2À iodide-bismuthate chains form an alternating stack that is weakly coupling, exhibiting aw eak 3D topological character. [9] Lately,alarge number of subhalides based on bismuth and transition metals have been discovered. [18][19][20] Theirs tructures typically consist of low-dimensional intermetallic fragments with strong bondingi nside and weaker,predominantly electrostatic interactions to their surroundings.L ayereds tructures comprising the same type of intermetallic kagome nets as Bi 14 Rh 3 I 9 are Bi 38 Pt 9 I 14 ,Bi 13 Pt 3 I 7 and Bi 12 Pt 3 I 5 . [13,21,22] Remarkably,t he intermetallic networks of several bismuth subhalides are tolerant to electronica lterations andf lexible enough to provide diffusion paths for excessive mass transports. For example, single-crystalso ft he metallics ubiodide Bi 13 Pt 3 I 7 were treated with n-butyllithium to obtain Bi 12 Pt 3 I 5 .I nterestingly,t he motif of the intermetallic network still persists even after such ah arsht reatment, demonstrating the stability of the net. [23] Similarly,a"breathing mode" of the intermetallic framework of Bi 12 Rh 3 Cl 2 allows as ubstitution of chloridei ons with bismuth resulting in the transformation into the binary intermetallic compound Bi 12 Rh 3 Bi 2 = Bi 14 Rh 3 . [19] The intermetallic layers of Bi 14 Rh 3 I 9 are most likely mechanically robust and not prone to distortion, rolling-up or furling. This opens al arge field for possible structurala nd chemical variations, like for example,d oping, de-/intercalation reactions and changes of stacking distances. Chemical gating [23] was suggested asaway to compensate the surfacep olarityt hat shifts the topological edge states away from the Fermi level in pristine samples. According to theoretical simulations based on Bi 14 Rh 3 I 9 ,t ransition metal (TM) exchange in the series of platinum group elements affects the size and shape of the topological band gap, but the topological character generally remains intact. [12] Only the 1 2 [(Bi 4 Rh) 3 I] 2 + intermetallic layer was considered in the calculation to reduce the structural complexity.I nt he case of TM = Rh, two non-trivial energy gaps can be found near the Fermienergy,h ence supporting the 2D TI state. Furthermore, 1 2 [(Bi 4 X) 3 I] 2 + layers with X = Ru, Pd, Os, Ir and Pt instead of Rh were studied theoretically,s howing topologically non-trivial energy gaps for Ir-, Pt-and Pd-based layers, while layers substituted with Os and Ru remaint rivial. This prediction has triggered the research interestinf urther weak 3D TIs.
Herein, we focus on compositional and structuralm anipulation of Bi 14 Rh 3 I 9 .I np articular,t he new phase Bi 12 Rh 3 Sn 3 I 9 is experimentallyo btained by formal substitution of bismuth by tin in the anionic spacer of Bi 14 Rh 3 I 9 .Bydoing so, central questions arise:D oes as ubstitution influence the band structure drastically?W ill the topological characterp ersist?I fs o, we might be able to achieve tuneable compositions andt opological band gaps. Moreover, this could open ap athway for the inclusiono f magnetic ions and thereby contributet ot he current quest for magnetic topological insulators. For ap ropera nswer to those questions, the electronic properties of Bi 12 Rh 3 Sn 3 I 9 were studied by density-functional theory (DFT) based calculations. Moreover,w ep erformed first temperature-dependent transport experimentsi no rder to underline the high conductivity within this layered material.

Results and Discussion
Thermal analysis of Bi 12 Rh 3 Sn 3 I 9 Insightful studies had already been devoteds olely to the synthesis of Bi 14 Rh 3 I 9 with respect to phase purity and crystal growth conditions. [17] According to the previous report on the ternary Bi-Rh-I system,D SC studies are provent ob ea ne ffective technique to attain deeper understanding of phasef ormation and thermal stability. Thus, investigations in the quaternary Bi-Rh-Sn-I system werep erformedb yh eating ag round mixture of Bi, Rh, Sn and BiI 3 in am olar ratio of 12:3:3:9 in af used silica ampuleu nder vacuum. The DSC data shows five significant signals (Figure 2), which can be addressed to various effects (Table1).  The signals were interpreted based on publishedd ata, [17] and by annealing stoichiometric mixtures of the startingm aterials for approximately 3h at the temperatures derived from the DSC analyses followed by an identification with X-ray powderd iffraction. The first endothermic effects, H1 at 232 8C and H2 at 271 8C, werea ssigned to the melting of the elements Sn [24] and Bi, [25] respectively.W hen heating the mixture iteratively to 280 8Ct wo weakly exothermic signals appear, which can be attributed to the formation of rt-Bi 2 Rh [26] and Bi 4 I 4 . [25] The latter decomposes at 292 8C( H3). In the next step, H4 at 305 8C, an unknown phase is formed, which was later identified as Bi 12 Rh 3 Sn 3 I 9 .O nce the temperature of about 475 8C( H5) is reached,t he decomposition of Bi 12 Rh 3 Sn 3 I 9 starts gradually.
After the DSC experimentst he resulting solid shows ad iffraction pattern similart oB i 14 Rh 3 I 9 .B y-products such as BiI 3 and Bi 4 I 4 are formed as well (Supporting Information Figure 1). Bi 12 Rh 3 Sn 3 I 9 is not perceptively sensitivet oa ir,b ut should be stored under inert gas.
Based on this study,amixture of Bi, Rh, Sn, and BiI 3 with the composition of the title compound was ground in ab all mill, pressed to ap ill, sintered at 310 8Cf or several days and subsequently quenchedt or oom temperature. Thist reatment lead to as ubstantial volumei ncreasea nd porosity of the pill. The chemicalc omposition was determined by EDX which confirmed the composition Bi 12 Rh 3 Sn 3 I 9 for microcrystalline powdera sw ell as single crystals, except for am inor tin deficiency ( Table 2).
The diffraction pattern of the product was very similart oa measured PXRD pattern of Bi 14 Rh 3 I 9 ( Figure 3), but the reflections were shiftedt owardl ower 2q angles. Using asilicon standard andt he knowncrystal structure of Bi 14 Rh 3 I 9 ,l attice parameters were determined by Rietveld refinement performed with TOPAS ( Figure 4; Supporting Information Table 2). To solve the crystal structureo ft he new compound, crystals were grown and investigated with single-crystal X-ray diffraction.

Crystals tructure of Bi 12 Rh 3 Sn 3 I 9
Considering the thermals tudies and the Ostwald-Miers range, an optimized temperature program was deduced for the growth of black hexagonal-shaped single-crystals of Bi 12 Rh 3 Sn 3 I 9 .S ince structuralp roblems, for example, twinning and stacking disorder similar as in Bi 14 Rh 3 I 9 , [17] can hamper the crystal structure refinements of the new compound, the synthesis was optimized to yield single-crystals up to 400 mm. Structure determinations using those crystals confirmed the composition Bi 12 Rh 3 Sn 3 I 9 and the substitution of bismuth(III) with tin(II) in the anionic spacer( Ta bles 3-5, Figure5).
The crystal structure of Bi 12 Rh 3 Sn 3 I 9 comprises intermetallic layers that are almost identicalt ot he kagomel ayers found in Bi 14 Rh 3 I 9 . [17] In contrast, ad ifferent anionic spacer is found in Bi 12 Rh 3 Sn 3 I 9 .I nB i 14 Rh 3 I 9 ,t he intermetallic layers are separated by iodido-bismuthate 1 1 [Bi 2 I 8 ] 2À zigzag chains,w hich can also    be seen as ad ouble layer of iodide ions in which half of the octahedral voids are filled by bismuth(III) cation. The sum formula can be structured as follows (with & representing av oid octahedral site) [Eq. (1)]: Bi 14 Rh 3 I 9 ¼fBi 12 Rh 3 Ig intermetallic net fBi 2 I 8 g spacer ¼ In contrast, the anionic spaceri nB i 12 Rh 3 Sn 3 I 9 is an iodidostannate layer The substitution of two bismuth(III) cations by three tin(II) cations is isoelectronic and does not change the layer charges. The assigned oxidation states are supported by the valence sums u of the cations, calculated based on bond-length bondstrength correlations: [27] u(Bi spacer ) = 2.99-3.04 for Bi 14 Rh 3 I 9 ,a nd u(Sn spacer ) = 1.89-1.99 for Bi 12 Rh 3 Sn 3 I 9 (Supporting Information Ta ble 3). The cross-check with opposite assignments yields no meaningful results(Supporting Information Ta ble 4).
In the crystal structure determination, the tin atoms appear to be disordered over all octahedral voids, yet not statistically with 75 %o ccupancyf or all position. Instead, the occupancies are rather different and specialized with occ(Sn1) = 83.6(1) % % 5/6 and occ(Sn2) = 66.4(1) % % 2/3. Ar easonable explanation is the following: The structure shows ah igh degree of pseudo-symmetry.T he layer symmetry of the intermetallic layer is P6/mmm. Within the accuracy of the lattice parameters, j c C cosß j /a = 1/6, that is, in orthogonal projectiono nt he (0 01)p lane,e very sixth unit cell exactly matches. Nonetheless, the symmetryo ft he structure is neither trigonal nor orthorhombic,b ecause the position of the symmetry elements of the individual layers do not match in space. Thus, al arger unit cell with c' = 6c C sinß is not appropriate.
The 1 2 [Sn 3 I 8 ] 2À spacerl ayer has ac orrugated surfacew ith I4 protruding by 0.29 from ap lane defined by I2 and I3. I4 is positioned directly above and below I1, which resides in the hexagonal prismatic voids of the intermetallic network and  corresponds to a" dent" in the surfaceo ft he latter (I1···I4 = 4.34 ). Thereby the two types of layers interlock. The most common ordered arrangement associatedw ith the average occupancy of 75 %i st he kagome net, which belongs to the uniform tilings. It is representedb yt he Schläfli symbol [3.6.3.6] and has the hexagonal plane-group symmetry P6mm. However,t his does not match the occupancies of 5/6 for Sn1 and 2/3 for Sn2. Ak agome-types pacerl ayer with long-range translational order along the stacking direction would correspond to occ(Sn1) = 100 %a nd occ(Sn2) = 50 %. Kagome-type spacerl ayers that are ordered within each layer but show statistical disorder alongt he stacking direction would correspond to occ(Sn1) = occ(Sn2) = 75 %. However,a nother cation distribution within the iodido-stannate layer is consistentw ith the refinement results:t he 2-isogonal tiling with Schläfli symbol [3.6.3.6;3 2 .6 2 ]a nd rectangularp lanegroup symmetry P2mm. Assuming ordered cation distribution in each layer but rotational disorder (1208 and 2408)i nt he sequenceo fs pacer layers,a verage occupancies of (100 % + 75 % + 75 %)/3 = 5/6 for Sn1 and (50 % + 75 % + 75 %)/3 = 2/3 for Sn2 result ( Figure 6). As this suggests al ocally ordered structure, we used the latter model for the electronic band structure calculations.T he according symmetry reduction, however,c ould not be resolved in the X-ray diffraction data because of missing long-range order (orientational stacking faults of the spacer layers).

Electronicstructure of Bi 12 Rh 3 Sn 3 I 9
The presence of crystallographic positions with partial occupancy as observed for the title compound requires appropriate structure models for the intended simulations. Here, two different models are applied: Model A:T he experimental structure as obtained from the present single-crystal XRD data (see Ta ble 3a nd Ta ble 4) is considered in this model,b ut with all crystallographic positions completely occupied. The observed partial occupancyo ft he two inequivalentS np ositions is modelled by appropriate reductiono ft he electron numbers using the virtual crystal approximation (VCA). Experimental occupancies are 83.6 %a nd 66.4 %f or Sn1 (4h)a nd Sn2 (4i), respectively,p roviding in the mean 1.67 and 1.33 valence electrons at the respective position. Using VCA, Sn1 is replaced by the tin-like pseudo-atom 49.67 Sn, and Sn2 by 49. 33 Sn. Here, the notation Z Sn meanst hat the pseudo-atom carries the nuclear charge Z and the same number of electrons,t hus guaranteeing overall chargen eutrality.T he advantage of this model is that it allows the use of the experimental crystal structure of Bi 12 Sn 3 Rh 3 I 9 .I ts disadvantage consists in ignoring any subtle effect of disordered vacancies at the Sn positions.
Model B:I nt his model, half of the Sn2 sites are left empty, according to a[ 3.6.3.6;3 2 .6 2 ]n et of tin(II) cations. All other Sn positions are completely occupied.I nt his way,t he model contains the correct number of Sn atoms and, thus, implies a decent description of interlayer bonding. The likely present local order of the material is described correctly but is supplemented by an ordered arrangemento ft he vacancies that is only indirectly supported by the X-ray data. This causes ar eductiono fc rystal symmetry,w hich is now described by the space group C1m1( no. 8). An advantage of ModelBis the possibility to account for local atomic relaxation, which is expectedt ob eo fi mportance in the neighborhood of the vacancy.T he structure details of Model Bw ithout and with relaxation are shown in the Supporting Information (Supporting Ta ble 5), where the latter was performed without any symmetry constraint (space group P1) but under preservation of size and shape of the unit cell. The largests hift, by about 0.25 ,i s observed for I7, which is one of the atomsn ext to the Sn-vacancy.
DFT calculations were performed for both described structure models.F igure 7s hows ac omparison of the total densities of states (DOS) of ModelAandModel B.
For the latter,o nly results obtained after relaxation of the atomic positions are shown herea nd in the following. The overall DOS of both models agree well in the whole presented energyr ange (main figure), while differences are visible around the top of valence band/bottom of conduction band (inset). In  The dominant contributions to the total DOS at and around the Fermi level (E F )o riginate from the Bi 6p and Rh 4d states of the 2D TI layer,a sh ad been found for Bi 14 Rh 3 I 9 .T he spacer layer provides essential contributionst ot he total DOS of the valence band only below À0.5 eV and to the conduction band above + 1.5 eV.
Related band structures are presented in Figure 8. We limit this presentation to ap seudo-hexagonal two-dimensional (2D) Brillouinz one in the plane spanned by the primitive lattice vectors a and b. This plane is parallel to the structural layers ( Figure 4). We denote the reciprocal lattice vectors of the 2D lattice generated by a and b as a*a nd b*. The pointsMand K, with M = a*/2 and K = (a* + b*)/3, resemble symmetry points of a2 Dh exagonal lattice but are not points of high symmetry in the present monoclinic (Model A) or triclinic (Model B) lattices. This is visible in Figure 8a tp oint K, where the dispersion shows small kinks.
As anticipated from the very similar DOS of both models, their band structureso nly differ in subtle detailsi nthe considered energy window close to the Fermi level. In particular, the four lowest conduction bands show almostt he same dispersion in both models. The bands of Model Aa re two-fold degenerate due to inversion symmetry,w hiler elatedb ands of Model Bs how small splitting. All bands close to the Fermi level show very small overall dispersion( < 0.5 eV).
We conclude this section by statingt hat the electronic structures provided by Model Aa nd Model Ba re qualitatively the same and both modelsu ndoubtedly yield an insulating ground state. The essential quantitative difference between both models consists in the size of the gap.

Z 2 invariants
For the considered structure model with centres of inversion, ModelA ,Z 2 invariants were calculated through Fu-Kane indices. [28] For the structure model without inversion centre, Model B, invariants werec alculated accordingt oR ef. [29].T his computationw as carried out directly from the PW92 band structure without resorting to an approximate Wannier representation. For both cases,w ef ind the invariants (n 0 ; n 1 , n 2 , n 3 ) = (0; 0, 0, 1), that is, the title compound is categorized as aw eak topological insulator with the same invariants as the parent compound Bi 14 Rh 3 I 9 . [9] The lowest gap within the conduction band carriest he same topological properties as the fundamental gap, while the second lowestg ap is trivial with (n 0 ; n 1 , n 2 , n 3 ) = (0;0 ,0 ,0 ). We confirmed the automatized calculation of these invariants for the case of Model Bb yv isual inspection of the Wannier centres.
The observed robustness of electronic structure with respect to structural details, together with the identification of identical sets of invariants in both models, provides confidencei n the validity of the structure models for the determination of the topological properties.

Transport measurements on Bi 12 Rh 3 Sn 3 I 9 crystals
In order to also investigate electronic properties of this new TI compound experimentally, we performed in situ four-point probe transportm easurements. The measurements were made under ultra-clean conditions, andt he controlled in situ positioningo ft ungsten tips as on-top contacts minimizes parasitic doping of the material. The linearity of the U I -curve clearly indicates metallicbehaviour (Figure 9).
The (spatially averaged) resistivity of the materialm easured at 300 Ki sa sl ow as 2.3 W & À1 and does not depend on the spacing d between the tips, which is indicative for a2 Dt ransport channel. The resistivity measured at 120 Ki ss lightly in-creased, but still metallic behaviour is obvious from the U Icurve, and thus activated transport along undoped bulk band channels plays only am inor role in this wide band gap TI material. Compared to previoust ransports tudies on Bi 2 Te 2 Se, [30] the conductivity is almost two orders of magnitude larger,d espite the weak band dispersion shown in DFT.W hether this is induced by ap ercolated network of protected 1D edge channels or by ap arasitic surface-neard oped bulk-band channel needs to be investigated in future experiments.

Conclusions
The exchange of bismuth(III) by tin(II) in the anionic spacer layer of Bi 14 Rh 3 I 9 leads to the new weak 3D TI Bi 12 Rh 3 Sn 3 I 9 . Thereby,t he iodido-bismuthate chains are replaced by iodidostannatel ayers [Sn 3 I 8 ] 2À while the intermetallic[ (Bi 4 Rh) 3 I] 2 + layers (2D TIs) are not subjectt oa ny modifications. DFT calculations also show aw eak 3D TI character of the novel tin-containing phase with at opological band gap of maximally 210 meV for ac ation-orderedm odel.T he tin substitution leads to as lightly enlarged molar volume, but decreases the size of the topological band gap compared to Bi 14 Rh 3 I 9 .N onetheless, layers with divalent cationsa re suitables pacers that efficiently prevente lectronic coupling between the 2D TI layers. At horough understanding of the crystal growth has benefited the synthesis of single crystals, whichm ay be used for further physicals tudies, for example, ARPES. Furthermore, the results hitherto have shown that the composition of Bi 14 Rh 3 I 9 can be altered not only theoretically, [12] but indeed experimentally.I n this respect, future research could emphasize on further modification of the anionic spacerb yd oping or intercalating other elements to generate new weak 3D TIs with adjusted band gaps. Of special interestw ould be to put magnetic cations in the spacer layer,w hich would be another step towards envisioned applications in low-energy spintronics and quantum computing.
Synthesis of Bi 12 Rh 3 Sn 3 I 9 powder and crystals:B ased on the results of thermal analyses, ap hase-pure microcrystalline powder of Bi 12 Rh 3 Sn 3 I 9 was synthesized by annealing as toichiometric mixture of bismuth, rhodium, and bismuth triiodide at 310 8Cf or five days. The starting materials were ground in ab all mill for about 25 minutes (Pulverisette 23, Fritsch), pressed to pellets (msscientific, diameter 6-8 mm) to ensure maximum homogeneity,a nd sealed in a3 mL evacuated silica ampule. After the heat treatment, the ampule was cooled to room temperature at ar ate of À5Kmin À1 . The growth of larger Bi 12 Rh 3 Sn 3 I 9 crystals was carried out in analogy to the procedure used for Bi 14 Rh 3 I 9 . [17] As toichiometric mixture of bismuth, rhodium, tin and bismuth triiodide was sealed in a3mL evacuated silica tube and heated from room temperature to 720 8C with ar ate of 12 Kmin À1 .T he sample was held at 720 8Cf or at least 10 minutes, followed by cooling to 420 8Ca tt he rate of À2Kmin À1 ,w here the temperature was kept for further 20 minutes. The process was continued by cooling to 310 8Ca tt he rate of À1Kh À1 and annealing at this temperature for approximately 7days. In the end, the ampule was quenched in cold water. Thermal analysis:D ifferential Scanning Calorimetry (DSC) was performed to investigate formation and decomposition processes in the quaternary Bi-Rh-Sn-I system. The starting materials in the desired ratio were sealed in at iny silica ampule. The measurements were conducted with this ampule under equilibrium pressure using aD TA-DSC Labsys TMA System (Setaram). The mixture was heated up to 800 8Ca nd cooled down to room temperature with a rate of 2o r5Kmin À1 . Powder and single-crystal X-ray diffraction:P owder X-ray diffraction (PXRD) data (Cu Ka 1 , l = 1.54059 , T = 296(1) K) were collected using either an X'Pert Pro diffractometer (PANalytical, Bragg-Brentano geometry,G e(2 20)h ybrid monochromator,f ixed divergence slits, PIXcel detector) or aS tadi Pd iffractometer (Stoe &C ie, Debye-Scherrer geometry,G e(111)m onochromator,D ectris Mythen 1K strip detector). The samples for PXRD measurements were ground and fixed on as ingle-crystal silicon sample holder. The Rietveld-refinement was performed with the program TOPAS [31,32] on am ixture of the target phase with Si standard. The references were extracted from the ICSD. Single-crystal X-ray diffraction (SCXRD) data were obtained using an Apex-II kappa CCD diffractometer (Bruker) or imaging plate diffractometer IPDS-II (STOE);M o Ka radiation, l = 0.71073 , T = 170(2) K. Numerical absorption corrections were applied based on optimized crystal descriptions. The structure was solved with direct methods and subsequent refinements against F o 2 .G raphics of the crystal structures were developed with Diamond. [33] Deposition Number 1989503 contains the supplementary crystallographic data for this paper. These data are provided free of charge by the joint Cambridge Crystallographic Data Centre and Fachinformationszentrum Karlsruhe Access Structures service www.ccdc.cam.ac.uk/structures.
Scanning electron microscopy and energy-dispersive X-ray spectroscopy:S canning electron microscopy (SEM) (U a = 5-15 kV) was performed using aS U8020 electron microscope (Hitachi)e quipped with multi detector system for secondary and low-energy backscattered electrons, while an Oxford Silicon Drift Detector X-Max N was used for the semi-quantitative energy-dispersive X-ray spectroscopy [34] (EDX) (U a = 20 kV). To acquire electron images, the powders were fixed on ac arbon pad settled on an aluminum sample holder.T oobtain the average composition by EDX, pellets of the powder (6-8 mm in diameter) were pressed and embedded into EpoThin TM2 epoxy resin (Buehler) and epoxy hardener (Buehler) under vacuum. After grinding the embedded pellets with aM eta-Serv 250 (Buehler,s ilicon carbide grinding paper) and subsequent polishing with aV ibroMet 2 (Buehler,M asterPrepTM alumina suspension), the surfaces were coated with carbon in an automatic rotary-pump coating system (Quorum Q150R ES). To collect EDX data of as ingle crystal, crystals were fixed directly on ac arbon film.
In-situ surface transport experiments:F our-point probe transport experiments were performed under ultra-high vacuum conditions at room temperature and 120 K( liquid nitrogen) by means of a four-tip scanning tunnelling microscope (STM) system (Omicron nanoprobe system) using NaOH-etched Wt ips with typical radii of 100 nm. Flakes of the samples (about 200 mmt hickness) were mounted on at ransferable sample plate. The Wt ips were navigated and positioned individually in the field of view of aS EM across the sample. This allows various probe geometries and defined probe spacings. The resistance values were corrected to calculate the conductivity of the sample. The resistivity was measured at various positions and probe currents (1-100 mA) in order to average out the effect of chemical inhomogeneities, which were partly seen in SEM. More details about surface sensitive four-point probe measurements can be found in Ref. [35]. Computational methods:D FT based calculations were performed with the full-potential local-orbital (FPLO) code, [36] version 18.00-52. [37] The exchange and correlation energy was considered in the local density approximation with PW92 parameterization. [38] Optionally,o ptimization of internal atomic coordinates was performed in scalar relativistic mode unless the forces on individual atoms fell below 20 meV À1 .T he final self-consistent calculations of the charge density were carried out using the four-component full-relativistic mode of FPLO. This effort is necessary due to the sizable spin-orbit coupling of all elements of the considered compound. The following basis states were treated as valence states (default FPLO basis set): Bi:5 s, 5p, 5d, 6s, 7s, 6p, 7p, 6d;R h: 4s, 4p, 5s, 6s, 4d, 5d, 5p;S na nd I: 4s, 4p, 4d, 5s, 6s, 5p, 6p, 5d. The k-space integrals were evaluated with the linear tetrahedron method using a k-mesh with 12 12 7i ntervals in the full Brillouin zone for all calculations.