# Fileset

[CoSn_sapphire_MDR.pdf](https://mdr.nims.go.jp/filesets/950c9b74-e663-4fe0-9981-97050dc4efe4/download)

## Creator

[Tomoya Nakatani](https://orcid.org/0000-0001-9590-216X), [Rohit Dahule](https://orcid.org/0000-0002-2768-8808), [Nattamon Suwannaharn](https://orcid.org/0000-0003-1285-599X), [Mevhibe Dilara Pütkül](https://orcid.org/0009-0007-0268-1478), [Yusuke Ikeda](https://orcid.org/0009-0001-4394-711X), [Taisuke T. Sasaki](https://orcid.org/0000-0002-5952-7638), [Ryoji Sahara](https://orcid.org/0000-0003-0788-2985)

## Rights

This article may be downloaded for personal use only. Any other use requires prior permission of the author and AIP Publishing. This article appeared in Tomoya Nakatani, Rohit Dahule, Nattamon Suwannaharn, Mevhibe Dilara Pütkül, Yusuke Ikeda, Taisuke T. Sasaki, Ryoji Sahara; Orientation-controlled epitaxy of CoSn anisotropic conductive thin films on sapphire substrates toward interconnect applications. J. Appl. Phys. 21 October 2026; 140 (15): 155101 and may be found at https://doi.org/10.1063/5.0352862.[In Copyright](http://rightsstatements.org/vocab/InC/1.0/)

## Other metadata

[Orientation-controlled epitaxy of CoSn anisotropic conductive thin films on sapphire substrates toward interconnect applications](https://mdr.nims.go.jp/datasets/ef978e03-03e9-47bb-8e99-eb91db04fdd0)

## Fulltext

Microsoft Word - CoSn\_on\_sapphire\_JAP\_rev\_CLEAN.docx  1  &#32; Orientation-controlled epitaxy of CoSn anisotropic conductive thin films on  sapphire substrates toward interconnect applications  &#32; Tomoya Nakatani,1\* Rohit Dahule,2 Nattamon Suwannaharn,2 Mevhibe Dilara Pütkül,1  Yusuke Ikeda,1 Taisuke T. Sasaki,2 and Ryoji Sahara,2  &#32; 1 Research Center for Magnetic and Spintronic Materials, National Institute for  Materials Science, 1-2-1, Sengen, Tsukuba, Ibaraki 305-0047, Japan.  2 Research Center for Structural Materials, National Institute for Materials Science, 1-2- 1, Sengen, Tsukuba, Ibaraki 305-0047, Japan.  &#32; \*Corresponding author’s email address: nakatani.tomoya@nims.go.jp  &#32;2  &#32; Abstract  The increasing resistivity of Cu interconnects at reduced device dimensions requires  alternative conductors with suppressed size effects of resistivity. Materials with highly  anisotropic electronic transport are a promising solution. CoSn exhibits pseudo-one- dimensional transport along the c axis with low resistivity comparable to conventional  metals, making it a promising candidate for next-generation interconnects. In this study,  we investigate the epitaxial growth of CoSn thin films on sapphire substrates with  controlled crystallographic orientations to align the low-resistivity c axis in the plane. We  successfully grow single-crystalline CoSn(1010) and CoSn(1120) films on sapphire  substrates using appropriate buffer layers. Although the (1120)  plane has a higher  atomic packing density and is thus expected to yield lower surface energy and smoother  films based on an analogy with conventional metals, the two orientations show markedly  different surface morphologies: CoSn (1010)  forms comparatively smooth films,  whereas CoSn(1120) exhibits pronounced {1010}-faceting and rough surfaces. First- principles calculations reveal that the (1010) surface has lower surface energy than the  (1120)  surface, consistent with the experimentally observed facet formation. This  demonstrates that surface energetics in CoSn are governed not by atomic packing density  but by anisotropic bonding associated with the kagome lattice. Electrical transport  measurements confirm strong dependence of resistivity on crystallographic orientation,  consistent with the bulk properties, with the lowest resistivity achieved along the c axis,  down to 10.6 μΩ cm for 50-nm-thick CoSn(1010). This work establishes orientation- controlled epitaxy on 300-mm-wafer-compatible sapphire substrates as a viable strategy  for integrating CoSn into next-generation interconnect architectures. 3  &#32; I. Introduction  As transistor dimensions continue to shrink, the electrical resistivity (ρ) of  conventional copper (Cu) interconnects increases significantly due to enhanced electron  scattering at surfaces and grain boundaries.1–3 This size effect originates from Cu’s nearly  isotropic Fermi surface and long electron mean free path (λ ≈ 39 nm at room  temperature).4 Consequently, electron scattering at surfaces and grain boundaries is  greatly increased in devices with small dimensions. To address this issue, materials with  a small 𝜌 𝜆 product (𝜌 : bulk resistivity) have been proposed, including platinum-group  elements such as Ru.5,6 Furthermore, materials exhibiting anisotropic electronic  conduction have emerged as promising candidates because their transport can be  engineered by crystallographic orientation.7,8 Candidate materials include delafossites9,10  and the CoSn kagome metal.7  CoSn crystallizes in a hexagonal B35 structure (hP6, space group P6/mmm), as  illustrated in Fig. 1(a), exhibiting highly anisotropic, pseudo-one-dimensional transport.  This property originates from its unique electronic band structure, which features a “flat  band” formed by the destructive interference of the wave function of conduction electrons  in the kagome lattice (c plane). While the resistivity in the c plane is high (\> 100 μΩ cm),  the resistivity is remarkably low along the \[0001\] c axis (3–7 µΩ cm in bulk 11–15), making  CoSn a promising candidate for next-generation interconnects.7 In our previous study of  single-crystalline CoSn(1010) films with an in-plane c axis, two critical challenges were  identified.16 First, the measured c axis resistivity of down to 13 µΩ cm remained higher  than the bulk value, suggesting additional scattering sources for conduction electrons,  such as defects or impurities. Second, the CoSn (1010)  films exhibited three-4  &#32; dimensional island growth, resulting in rough surfaces that are not suitable for device  integration. These issues highlight the need for improved epitaxial growth to realize  smooth and high-quality CoSn films.  For interconnect applications of CoSn, aligning the low-resistivity c axis parallel  to the longitudinal direction of the interconnect line and the current direction is essential.  Such alignment is difficult to achieve using polycrystalline films. To take advantage of  the anisotropic electronic conduction of CoSn in a multilevel interconnect structure, the  horizontal interconnects in each wiring layer must be patterned from a single-crystalline  CoSn film as straight lines parallel to its c axis. A potential solution is the “wafer transfer”  process, in which a single-crystalline CoSn film is grown on a template wafer and then  transferred to a device-fabricated wafer. Then, the template wafer is removed (debonded),  and the transferred CoSn film is patterned into interconnect lines. To implement this in  modern semiconductor manufacturing, epitaxial CoSn growth must be established on  300-mm-wafer-compatible substrates, which are currently limited to Si, SiC, and sapphire  (α-Al2O3).  Sapphire is a suitable substrate for the epitaxial growth of various materials and  the wafer transfer process due to its high mechanical, thermal, and chemical stability, as  well as its optical transparency to various wavelengths of light, including UV. Moreover,  the variety of available crystal planes in sapphire offers significant flexibility in epitaxial  design. In general, the choice of crystal plane strongly influences the epitaxial growth and  surface morphology of the material, which are linked to crystal symmetry and surface  energy. For growing single-crystalline CoSn films with an in-plane c axis, two epitaxial  planes are possible: the (1010) M-plane and the (1120) A-plane. As illustrated in Figs. 5  &#32; 1(b) and 1(c), the (1120) plane has a higher atomic packing density (10.3 atoms/nm2)  than the (1010) plane (8.9 atoms/nm2).   In simple metals with isotropic bonding, surface energy is well approximated by  the density of broken bonds and thus correlates with atomic packing density.17 Therefore,  densely packed planes generally have lower surface energy and tend to form flatter  surfaces in thin films. Based on this analogy, one may expect that CoSn(1120) epitaxy  would be favorable for achieving smoother films with an in-plane c axis compared to  CoSn(1010). However, CoSn is a kagome metal with intrinsically anisotropic bonding  associated with its anisotropic crystal structure. In such systems, surface energetics may  not be solely governed by atomic packing density because the number and nature of  broken bonds can vary significantly depending on the crystallographic orientation.  Therefore, a more detailed investigation of orientation-dependent growth and surface  morphology is required.  In this study, we investigate the epitaxial growth of CoSn single-crystalline films  on sapphire substrates. We demonstrate the growth of both CoSn(1010) and (1120)  films and systematically analyze how the crystallographic plane influences the surface  morphology, microstructure, and electronic transport properties. This study aims to  establish a viable pathway for integrating single-crystalline CoSn films into next- generation interconnect technologies.  &#32;6  &#32; &#32; FIG. 1. (a) Crystal structure of the kagome metal CoSn. (b) (1010) and (c) (1120)  planes of CoSn, representing the two candidate epitaxial planes for achieving in-plane c  axis orientation. (d) Atomic arrangement of the Ru (1010)  plane. (e) Atomic  arrangements of the Ru(1120) and Ti(1120), along with the oxygen sublattice of the  sapphire(1102)  R-plane. Schematic sample structures for (f) CoSn(1010)  and (g)  CoSn(1120) epitaxial growth.  &#32; II. Methods  (i) Experiments  All thin-film samples were deposited using magnetron sputtering in an Ar gas  atmosphere. The base pressure of the deposition chamber was \~2×10–6 Pa. As illustrated  in Figs. 1(b) and 1(d), the (1010) planes of CoSn and Ru exhibit good lattice matching, 7  &#32; with mismatch ratios of –0.5% along \[0001\] and –2.5% along \[1210\] relative to the  Ru lattice. Therefore, Ru(1010)  is a suitable template for growing CoSn(1010) . To  achieve this epitaxy, we used sapphire(1010) (M-plane) substrates and a complex buffer  architecture consisting of the following layers, from the substrate side: MgO(110) (5  nm)/Co50Fe50(211) (2 nm)/Co(1010) (2 nm)/Ru(1010) (2 nm), as shown in Fig. 1(f).  As reported by Talvacchio et al.18 and Stampe et al.,19 MgO(110) grows epitaxially on  sapphire(1010)  with an orientation relationship of \[110\] ∥ \[0001\]  . First,  the sapphire substrate was cleaned by heating it in the sputtering chamber at 600 °C for  10 min. The MgO layer was deposited by RF sputtering of a sintered MgO target at room  temperature (RT) and subsequently annealed at 600 °C for 10 min in the sputtering  chamber to improve the crystallinity of the MgO(110) single-crystalline film, which is  critical for the epitaxy of the overlying metallic layers. After depositing the Ru buffer  layer, a second annealing step was performed at 600 °C for 10 min, resulting in improved  crystal quality of the subsequent CoSn layer. The CoSn layers were co-sputtered from  elemental Co and Sn targets at a substrate temperature of 400 °C. The stoichiometry  (Co50Sn50 at. %) was precisely controlled within ±1% to ensure the phase purity of CoSn,  as previously described.16  For CoSn (1120)  films, sapphire (1102)  (R-plane) substrates were used.  Similar to the (1010) plane, the (1120) planes of Ru and CoSn exhibit good lattice  matching, as shown in Figs. 1(c) and (e). However, Ru(1120) could not be epitaxially  grown on sapphire (1102)  owing to the large lattice mismatch of –10.0% along  \[0001\]Ru and –8.2% along \[1100\]Ru between the oxygen sublattice of sapphire(1102)  and Ru(1120) . Therefore, we first deposited a Ti(1120)  (2 nm) layer at 400 °C on 8  &#32; sapphire(1102), exhibiting small lattice mismatch ratios of –1.6% along \[0001\]Ti and  +0.1% along \[1100\]Ti. Then, a Ru (4 nm) buffer layer was deposited at 400 °C. Despite  the relatively large lattice mismatch ratios of –8.6% and –8.3% along \[0001\]  and  \[1100\], respectively, the (1120)-plane epitaxy between Ti and Ru was obtained.  We characterized the crystal structure using X-ray diffraction (XRD) with Cu Kα  radiation and the microstructure using scanning transmission electron microscopy  (STEM). Surface morphologies were evaluated via atomic force microscopy (AFM). We  measured the resistivity of the CoSn epitaxial thin films using a four-probe method in  patterned devices with a width of 50 μm and a length of 100 μm, using a combination of  photolithography and Ar ion milling. As previously reported, resistance measurements in  unpatterned films using an in-line four-probe method yield inaccurate sheet resistance  values for materials with resistivity anisotropy.16  &#32; (ii) Computational details  First-principles calculations were performed within the framework of density  functional theory (DFT), as implemented in the Vienna Ab initio Simulation Package  (VASP).20,21 The Perdew-Burke-Ernzerhof generalized gradient approximation (PBE- GGA) was adopted for the exchange-correlation functional,22 together with the projector  augmented-wave (PAW) method.23 The valence electron configurations considered were  3d84s1 for Co and 5s25p2 for Sn. A plane-wave kinetic energy cutoff of 450 eV was used  for all calculations. The bulk structure was fully optimized prior to surface construction  using a Γ-centered 10×10×11 Monkhorst-Pack k-point mesh for Brillouin zone  integration.24 The optimized lattice parameters of CoSn are a = 0.529 nm and c = 0.422 9  &#32; nm, which are close to the experimental values (a = 0.528 nm and c = 0.426 nm).25  Electronic self-consistency was achieved with a convergence criterion of 10–6 eV, and the  force convergence threshold was set to 0.01 eV/Å.  Surface slabs for the (0001), (1010)  and (1120)  orientations were  constructed from the optimized bulk structure. Each slab naturally exposes two opposing  planes, (0001) and (0002), (1010)  and (2020) , and (1120)  and (2240)  arising  from the bulk stacking sequence. A five-layer slab geometry was employed along the  surface normal, separated by a vacuum region of 15 Å to prevent spurious interactions  between periodically repeated images (see Fig. S3). The atoms in the bottom two layers  were fixed to mimic bulk-like behavior, while the upper three layers were allowed to fully  relax. Dipole corrections were applied along the surface normal to eliminate artificial  electrostatic interactions.26 Brillouin zone integration for the surface calculations was  performed using Γ-centered k-point meshes of 10×10×1 for (0001), (1010)  surfaces  and 6×10×1 for (1120) surface.   Surface energies were evaluated using a two-step approach involving cleavage  and relaxation contributions.27 The cleavage energy, 𝐸 , was calculated from the  unrelaxed slab as   𝐸 = \[𝐸 − 𝑁 × 𝐸 \],  (1)  where 𝐸  is the total energy of the unrelaxed slab, 𝐸  is the energy of the bulk  reference structure, 𝑁 is the number of bulk formula units contained in the slab and A is  surface area. Structural relaxation of the upper layers then provided the relaxation energy,  𝐸 ,  𝐸 = \[𝐸 − 𝐸 \],   (2) 10  &#32; where 𝐸  is the total energy of the relaxed slab. The total surface energy, 𝛶 , is  obtained as  𝛶 =  𝐸 + 𝐸 .   (3)  &#32; III. Results  A. Crystal structures and surface morphology of CoSn(𝟏𝟎𝟏𝟎)  First, we describe the growth of CoSn(1010)  on the sapphire(1010)  via the  MgO(110)/CoFe(211)/Co(1010) /Ru(1010)  buffer system. For comparison, the same  buffer layers and CoSn films were deposited on an MgO(110) substrate. Figure 2(a)  shows the out-of-plane XRD profiles of a 10-nm-thick MgO film on sapphire(1010) in  both as-deposited and annealed states. In both cases, only the MgO 220 peak was  observed, confirming the out-of-plane alignment of the (110)-plane of MgO. The  annealing process at 600 °C clearly increased the peak intensity, indicating improved  MgO crystallinity. The inset of Fig. 2(a) shows the 𝜙 -scan profile of the MgO 200  reflections at 2θ = 42.9° with a χ tilt of 45°; the twofold symmetry confirms that  MgO(110) grows as a single crystal with the orientation relationship of \[110\] ∥ \[0001\]  as reported previously.18,19  Figure 2(b) shows the out-of-plane XRD profiles for 50-nm-thick CoSn films  grown on both MgO(110) and sapphire(1010)  substrates via MgO (5 nm)/CoFe (2  nm)/Co (2 nm)/Ru (2 nm) buffer layers. The presence of the CoSn 1010 reflection and  its higher-order peaks indicates a preferred out-of-plane alignment of CoSn(1010) .  However, as shown in Fig. 2(b), the CoSn 0002 peak at 2θ = 42.0° was observed only for  the film on the sapphire substrate, suggesting the presence of misoriented grains in the 11  &#32; film. Figure 2(c) shows the 𝜙-scan profiles of the CoSn 1120 reflections, where the  twofold symmetry confirms the single-crystalline growth of CoSn with the relationship  of \[0001\] ∥ \[110\] ∥ \[1210\] . The tCoSn-dependence of the full width at  half maximum (FWHM) for the rocking curves of the CoSn 1010 reflection along the  β-axis \[inset of Fig. 2(b)\] and 𝜙-axis (Fig. 2(c)\]is summarized in Fig. 2(d). The β-axis  corresponds to the azimuthal coordinate of the diffraction spot on the two-dimensional  detector, where β = 180° is defined as the surface normal direction. Therefore, the FWHM  along the β-axis reflects the tilt distribution of the corresponding crystallographic planes,  i.e., the CoSn(1010)  planes in the present case. For the same sample, the FWHM  measured along the β-axis is approximately 0.8° larger than that measured along the ω- axis due to the different optical configurations employed in these measurements.  Therefore, the β-axis FWHMs in Figs. 2(d) and 3(d) are used to compare the relative  evolution of crystalline alignment among the CoSn films. Although the FWHM values  decrease with increasing tCoSn for both substrates, the crystallinity of the CoSn films on  sapphire substrates remains inferior to that on MgO(110) at any given thickness.  The surface morphology of the CoSn films was characterized by AFM, as shown  in Figs. 2(e) and 2(f). For tCoSn = 10 nm, the CoSn surface exhibits a three-dimensional  island-like morphology (Fig. 2(e)). The peak-to-valley (p-v) values exceeded 20 nm,  which is greater than the nominal film thickness of 10 nm, indicating that the islands are  at least partially disconnected. For tCoSn = 50 nm (Fig. 2(f)), the arithmetic average  roughness (Ra) decreased, and the p-v values became smaller than the film thickness,  suggesting the formation of a continuous film. Furthermore, the AFM images reveal that  the surface grains are elongated along the c axis, suggesting a higher crystal growth 12  &#32; velocity along the c axis than the a axis.  &#32; &#32; &#32; &#32; FIG. 2. Epitaxial growth of CoSn(1010). (a) Out-of-plane XRD profiles of MgO (10  nm) on a sapphire(1010) substrate. The inset shows 𝜙-scan profile from the MgO 200  reflections. (b) Out-of-plane XRD profiles of MgO (5 nm)/CoFe (2 nm)/Co (2 nm)/Ru (2  nm)/CoSn (50 nm) films on MgO(110) and sapphire (1010)  substrates. The inset  displays the rocking curve of CoSn 1010 peak along the β-axis. (c) 𝜙-scan profile of  the CoSn 1120  reflection. (d) tCoSn dependence of the full width at half maximum  (FWHM) of the rocking curves for the β and 𝜙 axes. AFM surface images of the CoSn  films with thicknesses of (e) tCoSn = 10 nm and (f) tCoSn = 50 nm on MgO(110) and  sapphire(1010) substrates.  &#32; B. Crystal structures and surface morphology of CoSn(𝟏𝟏𝟐𝟎) 13  &#32; Next, we discuss the growth of CoSn(1120) on the sapphire(1102) substrate/Ti  (2 nm)/Ru (4 nm) buffer system. Figure 3(a) shows a two-dimensional XRD image of a  50-nm-thick CoSn sample. As shown in Fig. 3(a) and the inset of Fig. 3(b), the CoSn  1120  and 2240  peaks, as well as the Ru 1120  peak, are shifted from the surface  normal (β = 180°) by approximately 1°. This indicates a tilt of the (1120) planes of Ru  and CoSn relative to the sapphire (1102)  substrate. The Ti 1120  peak, which is  expected to appear at 2θ = 62.6°, was not observed because the 2-nm-thick Ti layer was  too thin to exhibit sufficient XRD intensity. To investigate this further, we prepared a  separate 10-nm-thick Ti film on sapphire(1102). The Ti 1120 peak appeared exactly at  β = 180° with a narrow FWHM of 0.7°, significantly smaller than that of Ru (FWHM =  2\.9°) \[inset of Fig. 3(b)\]. These results confirm that Ti (1120)  grows with high  crystallinity parallel to the substrate surface, as expected from the lattice matching  between sapphire(1102)  and Ti(1120)  (Fig. 1(e)). Therefore, the (1120)  planes of  Ru and CoSn are tilted by approximately 1° with respect to the underlying Ti(1120)  plane, most likely due to the 8.6% lattice mismatch between Ti and Ru.  Figures 3(b) and 3(c) show the out-of-plane XRD profiles and the 𝜙-scan profiles  of the CoSn 1010 reflections, respectively. The twofold symmetry observed in the 𝜙- scan confirms the single-crystalline growth of CoSn(1120). As shown in Fig. 3(d), the  crystallinity of CoSn improves with increasing tCoSn, similar to the trend observed in the  (1010) films \[Fig. 2(d)\]. However, the surface morphology remained significantly rough.  Figure 3(e) shows the AFM image and height (z-scale) line profile measured  perpendicular to the c axis for tCoSn = 10 nm. This reveals a highly directional morphology  with sharp ridges oriented along the c axis, indicating a higher crystal growth velocity 14  &#32; along the c axis and pronounced facet growth. For tCoSn = 50 nm \[Fig. 3(f)\], the surface  exhibits flat terraces and deep voids with a p-v value of 71.4 nm, which exceeds the  nominal film thickness. Contrary to our initial expectation based on the atomic packing  density of CoSn(1010)  and (1120)  \[Figs. 1(b) and 1(c)\], the epitaxial growth of  CoSn(1120) failed to yield a flat surface and instead resulted in a rougher morphology  than that of CoSn(1010).  &#32; &#32; FIG. 3. Epitaxial growth of CoSn(1120). (a) Two-dimensional (2D) XRD image of a  CoSn(1120) (tCoSn = 50 nm) film on sapphire(1102)/Ti (2 nm)/Ru (4 nm). (b) Out-of- plane XRD profiles for tCoSn = 10–50 nm, integrated from the 2D images over β = 175°– 185°. The inset shows rocking curves of Ru 1120 and CoSn 2240 reflections along  the β-axis. The rocking curve of the Ti 1120  reflection from Ti (10 nm) on  sapphire(1102) is also shown. (c) 𝜙-scan profile from the CoSn 1010 reflection. (d)  tCoSn dependence of the FWHM of the rocking curves for the β and 𝜙 axes. AFM surface 15  &#32; images and height line profiles measured perpendicular to the c axis for (e) tCoSn = 10 nm  and (f) tCoSn = 50 nm.  &#32; C. Microstructure  Figures 4(a) and 4(b) show low-magnification high-angle annular dark-field  (HAADF)-STEM images of CoSn(1010) films with tCoSn = 10 and 50 nm grown on  sapphire(1010) substrates via MgO (5 nm)/CoFe (2 nm)/Co (2 nm)/Ru (2 nm) buffer  layers. The images were taken along the \[0001\] zone axis (c axis) of CoSn. Since the  contrast in HAADF-STEM images is approximately proportional to the square of the  atomic number (𝑍 ), the bright thin layer corresponds to the 2-nm-thick Ru (𝑍 = 44)  buffer layer. Although the Ru buffer layer is relatively flat, the 10-nm-thick CoSn film  exhibits a rough surface morphology, while the 50-nm-thick CoSn film appears relatively  smooth. These observations are consistent with our previous report on CoSn(1010)  grown on MgO(110) substrates,16 as well as with the reduced surface roughness observed  in thicker samples via AFM measurements \[Figs. 2(e) and 2(f)\]. However, the Ru buffer  layer appears faint or discontinuous in some regions, as indicated by an arrow in Fig. 4(b),  although the CoSn(1010) growth is still observed on top of these regions.  Figures 4(c) and 4(d) show the overlaid energy dispersive X-ray spectroscopy  (EDS) elemental concentration maps of O, Mg, Co, Ru, and Pt (protective layer), together  with the corresponding line profiles for the 10-nm-thick CoSn sample. In both samples,  an approximately 3-nm-thick oxidized layer was observed on top of the CoSn film. The  Co and Sn distributions within the CoSn layer remained relatively uniform and  homogeneous, with no clear evidence of compositional fluctuation observed inside the 16  &#32; CoSn film. The broadening of the Ru concentration toward the CoFe buffer layer  observed in the concentration line profile suggests that Ru has diffused partially into the  Co and CoFe buffer layers, which may be related to the solid solubility of Ru and Co.  This diffusion may also be related to the observed Fe segregation at the MgO/CoFe and  Ru/CoSn interfaces. However, Ru diffusion into the CoSn layer was not detected. Similar  EDS characteristics were also observed for the 50-nm-thick CoSn film. Oxygen was also  detected inside the CoSn layer at around 20 at. %. Considering the presence of a 3-nm- thick oxidized layer on CoSn surface, similar passivated oxide layers may also form on  the front and back surfaces of the TEM lamella after specimen preparation. Therefore,  the measured oxygen concentration in the CoSn layer may be overestimated due to these  passivation layers. X-ray photoelectron spectroscopy (XPS) measurements indicated that  no significant oxygen signal was detected in the interior of the CoSn layer. \[see Sec. 2 of  the Supplementary Material\]  &#32; &#32;17  &#32; FIG. 4. HAADF-STEM images (a, b) and EDS elemental maps (c, d) of CoSn(1010)   films grown with nominal thicknesses of 10 and 50 nm, respectively, together with the  corresponding EDS line profile for the 10-nm sample.  &#32; &#32; FIG. 5. HAADF-STEM images (a, b) and EDS elemental maps (c, d) of CoSn(1120)  films grown with thicknesses of 10 and 50 nm, respectively, together with the  corresponding EDS line profiles for the 50-nm sample.  &#32; Figures 5(a) and 5(b) show HAADF-STEM images of CoSn(1120) films with  tCoSn = 10 and 50 nm, respectively, grown on sapphire(1102) substrates via Ti (2 nm)/Ru  (4 nm) buffer layers. These images were taken along the \[0001\] zone axis of CoSn. The  10-nm-thick CoSn (1120)  film exhibited distinct facet growth, with isolated CoSn  islands showing {1010} facets. This is consistent with the AFM image, as shown in Fig.  3(e). In contrast, as the film thickness increased to 50 nm, the CoSn islands coalesced, 18  &#32; partially forming the designed (1120) top facet. The bright contrast corresponding to  the thin Ru buffer layer in the tCoSn = 10 nm sample appeared relatively flat and continuous  within the observed region. However, in the tCoSn = 50 nm sample, the Ru buffer layer  became discontinuous or faint in some regions.  EDS elemental maps of the 10-nm and 50-nm CoSn samples are shown in Figs.  5(c) and 5(d), respectively. While the elemental distributions of Co and Sn appeared  relatively uniform in the 10-nm sample \[Fig. 5(c)\], elemental diffusion and compositional  inhomogeneity were observed in the 50-nm sample \[Fig. 5(d)\]. Compared to region (A),  region (B) exhibited Co depletion and Sn enrichment in areas above the Ru buffer layer.  Ru diffusion into the CoSn film was also detected in these regions, which may be related  to the solid solubility of Co and Ru, as shown in the EDS line profiles. Therefore,  interdiffusion occurred in regions where the Ru layer was discontinuous.  As shown in Fig. 6 (a) and 6(b), high-magnification HAADF-STEM images clearly  reveal the kagome structure along the \[0001\] zone axis for the 50-nm-thick CoSn(1010)  and CoSn(1120) films, respectively. Despite their different growth orientations, domain  boundaries were observed in both films. These domain boundaries exhibited similar  structures and were aligned parallel to the (1010)  plane, similar to those previously  reported for CoSn(1010) films grown on MgO(110) substrates.16 The identification of  the structure of the domain boundaries requires STEM observations along another zone- axis and image-simulation studies. Therefore, detailed structural characterization of the  domain boundaries and evaluation of their effect on electronic transport properties remain  subjects for future study.  &#32;19  &#32; &#32; FIG. 6. High-magnification HAADF-STEM images of 50-nm-thick (a) CoSn(1010)  and (b) CoSn(1120) films.  &#32; &#32; D. Anisotropic resistivity  The resistivity of the CoSn films (𝜌 ) was obtained as 𝜌 = 𝑡 /(1/𝑅 − 1/𝑅 ) , where 𝑅   and 𝑅   are the sheet resistances of the entire film and  buffer layers, respectively, and 𝑡  is the nominal CoSn thickness determined from the  pre-calibrated deposition rate and deposition time. This nominal thickness corresponds to  the average film thickness for films with rough surface morphology. The method for  estimating 𝑅  is described in the Supplementary Material. When calculating 𝜌 ,  we considered the errors of 𝑅 , 𝑅 , and 𝑡 . The combined error by 𝑅   and 𝑡   was 1.7%, while that of 𝑅   was much larger (see the Supplementary  Material). The in-plane directions orthogonal to the low-resistivity c axis are \[1210\] (a  axis) and \[1100\]  for the CoSn(1010)  and CoSn(1120)  films, respectively. In ideal 20  &#32; hexagonal crystal, hexagonal symmetry requires isotropic resistivity within the basal  plane,28 therefore, 𝜌∥ = 𝜌∥\[ \] = 𝜌∥ .  Figure 7(a) shows the resistivity of CoSn(1010)  films with various tCoSn values  grown on MgO(110) and sapphire(1010)  substrates. For both substrates, the 𝜌∥   values were 100–140 μΩ cm for tCoSn = 10–50 nm, consistent with the values in the bulk  single crystals ( 𝜌∥   \~120 μΩ cm).13,15 The c axis resistivity (𝜌∥ )  exhibited  similar values for the MgO(110) and sapphire(1010) substrates. As shown in Fig. 2(d),  the CoSn films grown on the sapphire(1010) substrate exhibited lower crystallinity than  those grown on the MgO(110) substrate. However, the similar 𝜌∥  values on these  substrates indicate that this degree of difference in the (1010) -crystallinity is not  significant for 𝜌∥ . The value of 𝜌∥  depends strongly on tCoSn: 𝜌∥  = 10.6 μΩ  cm for tCoSn = 50 nm and 𝜌∥  \~30 μΩ cm for tCoSn = 10 nm. However, since the CoSn  surface roughness is more significant in thinner films due to the three-dimensional  CoSn (1010)  growth, the observed 𝜌∥  –tCoSn relationship contains an extrinsic  contribution from surface roughness. Therefore, we cannot evaluate the intrinsic  dependence of 𝜌∥  on tCoSn unless we achieve smooth CoSn film surfaces.  Figure 7(b) shows the resistivity of CoSn(1120) films with different tCoSn grown  on a sapphire(1102) substrate using Ti (2 nm)/Ru (4 nm) buffer layers. Similar to the  CoSn(1010) films, 𝜌∥  increased with decreasing tCoSn. However, at a given tCoSn,  𝜌∥   was higher for the CoSn(1120)  films than for the CoSn(1010)  films; for  example, 𝜌∥  = 17.1 and 10.6 μΩ cm for 50-nm-thick CoSn(1120) and CoSn(1010),  respectively. Similarly, at a given tCoSn, 𝜌∥  was higher for the CoSn(1120) films 21  &#32; than for the CoSn(1010) films. These differences in resistivity between the two epitaxial  orientations may arise from several structural factors. Domain boundaries with similar  structures and apparently comparable densities within the observed regions were found  in both films, as shown in Fig. 6. In addition, the CoSn(1010) and CoSn(1120) films  exhibited comparable rocking-curve FWHMs at a given tCoSn, as shown in Figs. 2(d) and  3(d). These observations suggest that neither domain-boundary scattering nor differences  in the degree of crystalline alignment are likely to be the primary origin of the systematic  resistivity difference.   In contrast, interdiffusion between Ru and CoSn was detected in the 50-nm-thick  CoSn(1120) film but not in the CoSn(1010) films. This interdiffusion, together with  the associated compositional inhomogeneity, may contribute to the higher resistivity of  the 50-nm-thick CoSn (1120)  film. However, because such interdiffusion was not  detected in the 10-nm-thick CoSn(1120) film within the observation area for STEM, its  contribution to the systematic resistivity difference between these epitaxial orientations  cannot be determined from the present results.  The pronounced surface roughness of the CoSn(1120) films due to the {1010}  facet growth may also contribute to the higher resistivity. In particular, the 10-nm-thick  CoSn(1120) consisted of {1010}-faceted islands elongated along the c axis \[Figs. 3(e)  and 5(a)\]. This surface morphology may account for 𝜌∥  of \~206 μΩ cm, which is  significantly higher than that for the 10-nm-thick CoSn(1010) film (\~110 μΩ cm). In  addition, the 10-nm-thick CoSn (1120)  film exhibited nearly isolated islands with  limited connectivity along the \[1100\] direction, whereas substantial island coalescence  was observed in the 50-nm-thick films (Fig. 5). This thickness-dependent improvement 22  &#32; in film connectivity may partly explain the increasing 𝜌∥  with decreasing tCoSn for  the CoSn(1120) films.  &#32; &#32; &#32; FIG. 7. tCoSn-dependence of the resistivity of (a) CoSn(1010)  and (b) CoSn(1120)  films along the c axis and in the basal plane. The effect of the current shunting through  the buffer layers has been subtracted.  &#32; IV. Discussion  The experimentally observed difference in surface morphology between the  CoSn (1010)  and (1120)  single-crystalline films is counterintuitive from the  viewpoint of conventional metallic thin films. In simple metals with nearly isotropic  bonding, densely packed crystallographic planes generally exhibit lower surface energies  and therefore tend to form flatter films. Because the CoSn(1120) plane has a higher  atomic packing density (10.3 atoms/nm2) than the CoSn(1010) plane (8.9 atoms/nm2)  \[Figs. 1(b) and 1(c)\], one would expect CoSn(1120)  epitaxy to produce smoother  10 20 30 40 50 0 20 40 100 120 140 160 r (m W &#32;c m ) tCoSn (nm) Black: MgO(110) Red: sapphire M-plane ∥basal plane ∥c axis (a) CoSn(1010) 10 20 30 40 50 0 20 40 160 180 200 220 240 r (m W &#32;c m ) tCoSn (nm) ∥c axis ∥basal plane (b) CoSn(1120) 206+164 -6423  &#32; surfaces. However, our experiments revealed the opposite trend: CoSn(1010)  forms  comparatively smooth films, whereas CoSn(1120)  exhibits pronounced faceting and  rough surfaces. This discrepancy clearly indicates that atomic packing density alone is  insufficient to describe the surface stability in CoSn.  To clarify the origin of the observed surface morphology, we calculated the  cleavage, relaxation, and total surface energies of representative CoSn surfaces using DFT.  The results are summarized in Table 1. Except for the (2240) surface, the relaxation  energy associated with surface reconstruction did not significantly impact the surface  energy \[see Sec. 3 of the Supplementary Material\]. The surface energy follows the  sequence of (0001) \< (2020) \< (0002) \< (1010) \< (2240) \< (1120). The (0001)  surface (kagome plane) exhibits both the highest atomic packing density (14.3  atoms/nm2) and the lowest surface energy. However, this correlation between atomic  packing density and surface energy breaks down when comparing (1010) and (1120)  surfaces, as well as (2020)  and (2240)  surfaces. This behavior suggests that the  cleavage energy is primarily governed by the number and topology of broken Co-Sn, Co- Co, and Sn-Sn bonds associated with the kagome network rather than by atomic packing  density alone. Consequently, crystallographic planes with relatively high atomic packing  density do not necessarily have low surface energies in CoSn. Additional analyses of  surface relaxation and charge redistribution are presented in the Supplementary Material.  It should be noted that surface energies alone do not determine the growth mode  and surface morphology of sputtered CoSn thin films. Thin-film morphology can also be  influenced by kinetic factors, such as adatom mobility and anisotropic growth rates, as  well as by interface energy and elastic strain energy. Therefore, the present DFT results 24  &#32; should be interpreted as explaining the preferential expression of low-energy facets within  the orientation-constrained epitaxial geometry, rather than providing a complete  description of the growth mode of CoSn thin films.   Although the CoSn(0001) surface has the lowest surface energy, it is not the  relevant top-surface orientation for the present films, in which either CoSn(1010) or  CoSn(1120) is constrained to be parallel to the substrate surface, thereby aligning the c  axis within the film plane. In addition, the elongated surface morphology observed by  AFM suggests that crystal growth is faster along the c axis than along the transverse in- plane direction. Under this orientation constraint, the relevant comparison is between  non-basal surfaces parallel to the c axis, namely, CoSn (1010)  and CoSn (1120)  surfaces.  &#32;At the elevated growth temperature of 400 °C employed in this study, sputtered  adatoms are expected to have high surface mobility, which can promote aggregation and  three-dimensional island growth. Once such island growth occurs, CoSn islands are  expected to preferentially expose {1010} facets because the calculated surface energy  of CoSn(1010) is lower than that of CoSn(1120). This tendency is consistent with the  preferential appearance of {1010} facets observed experimentally in both CoSn(1010)  and CoSn(1120) epitaxial films. With increasing thickness, these islands coalesce into  a continuous film, resulting in a substantially smoother morphology, as observed for the  CoSn(1010) film with tCoSn = 50 nm \[Fig. 4(b)\]. These results suggest that improving  thin-film smoothness further will require suppressing facet-mediated island growth while  maintaining the favorable CoSn(1010) orientation that aligns the low-resistivity c axis  within the film plane. 25  &#32; Table 1 Cleavage, relaxation and total surface energies (in eV Å–2) of CoSn surfaces.    Surface Ecleavage Erelaxation 𝜰𝐬𝐮𝐫𝐟𝐚𝐜𝐞  (0001) 0.080 –0.008 0.072  (0002) 0.080 –0.001 0.079  (10𝟏0) 0.087 –0.006 0.081  (20𝟐0) 0.087 –0.009 0.078  (11𝟐0) 0.105 –0.005 0.100  (22𝟒0) 0.105 –0.017 0.088  &#32; &#32; V. Conclusions  Single-crystalline CoSn (1010)  and (1120)  films were grown on  sapphire (1010)  and (1102)  substrates using appropriate buffer architectures.  CoSn(1010) films initially exhibit three-dimensional island growth, leading to a rough  surface at small thicknesses. With increasing thickness, the islands coalesce into a  continuous film with improved surface flatness. A 50-nm-thick CoSn (1010)  film  exhibits pronounced resistivity anisotropy, with 𝜌∥  = 10.6 μΩ cm and 𝜌∥  = 120  μΩ cm, approaching the intrinsic anisotropic transport properties of bulk CoSn. In  contrast, CoSn(1120) films exhibit pronounced facet formation with {1010} planes,  resulting in significantly rougher surfaces and higher resistivity. First-principles  calculations reveal that the (1010) surface has a lower surface energy than the (1120)  surface. This explains the faceted growth behavior despite the higher atomic packing  density of the (1120)  surface. These results demonstrate that the surface stability in  CoSn is not governed by atomic packing density but rather by the anisotropic bonding  characteristics of the kagome lattice. 26  &#32; These findings demonstrate that CoSn (1010)  epitaxy is the most suitable  approach for achieving relatively smooth, single-crystalline films with the in-plane c axis  orientation required for interconnect applications. This work provides design guidelines  for the orientation-controlled growth of anisotropic conductors and emphasizes the  critical role of surface energetics in determining film morphology and transport  performance. To implement CoSn interconnects, future work should establish full-wafer  epitaxy on 300-mm-compatible substrates together with a suitable sacrificial layer for  wafer debonding, and develop compatible chemical mechanical polishing, wafer-bonding,  and CoSn etching processes.  &#32; Acknowledgements  &#32;We acknowledge Kosuke Imamura and Mitsuru Ohtake (Yokohama National  University) for valuable discussions. This work was supported by "JST-Research and  Development Program for Next-generation Edge AI Semiconductors” Grant Number  JPMJES2531. The XRD and XPS measurements were supported by ARIM of MEXT  (JPMXP1225NM5220 and JPMXP1226NM5145). The computational results presented  in this work were obtained using the Numerical Materials Simulator supercomputing  facilities at NIMS.  &#32; Supplementary Material  Refer to the Supplementary Material for the method to estimate the buffer-layer sheet  resistance, XPS data, and additional analyses of surface relaxation and charge  redistribution based on DFT calculations. 27  &#32; Data Availability Statement  The data supporting the findings of this study are available from the corresponding author  upon reasonable request.  &#32; Conflict of interest  The authors have no conflict of interest regarding the publication of this article.  &#32; References  1 D. Gall, J.J. Cha, Z. Chen, H.J. Han, C. Hinkle, J.A. Robinson, R.  Sundararaman, and R. Torsi, “Materials for interconnects,” MRS Bull. 46(10),  959–966 (2021).  2 J.-P. Soulié, K. Sankaran, B. Van Troeye, A. Leśniewska, O. Varela Pedreira, H.  Oprins, G. Delie, C. Fleischmann, L. Boakes, C. Rolin, L.-Å. Ragnarsson, K.  Croes, S. Park, J. Swerts, G. Pourtois, Z. Tőkei, and C. Adelmann, “Selecting  alternative metals for advanced interconnects,” J. Appl. Phys. 136, 171101  (2024).  3 J.S. Kim, J. Kim, D.J. Yang, J. Shim, L. Hu, C.S. Lee, J. Kim, and S.W. Kim,  “Addressing interconnect challenges for enhanced computing performance,”  Science 386(6727), eadk6189 (2024).  4 J.S. Chawla, and D. Gall, “Specular electron scattering at single-crystal Cu(001)  surfaces,” Appl. Phys. Lett. 94, 252101 (2009).  5 D. Gall, “Electron mean free path in elemental metals,” J. Appl. Phys. 119,  085101 (2016).  6 S. Dutta, K. Sankaran, K. Moors, G. Pourtois, S. Van Elshocht, J. Bömmels, W.  Vandervorst, Z. Tokei, and C. Adelmann, “Thickness dependence of the  resistivity of platinum-group metal thin films,” J. Appl. Phys. 122, 025107  (2017).  7 S. Kumar, C. Multunas, B. Defay, D. Gall, and R. Sundararaman, “Ultralow  electron-surface scattering in nanoscale metals leveraging Fermi-surface  anisotropy,” Phys. Rev. Mater. 6, 085002 (2022).  8 K. Moors, K. Sankaran, G. Pourtois, and C. Adelmann, “First-principles-based  screening method for resistivity scaling of anisotropic metals,” Phys. Rev. Mater. 28  &#32; 6(12), 123804 (2022).  9 Y. Li, G. Zhou, M.M. Kelley, S. Shahriyar Nishat, S. Bey, M. Abdul Karim, X.  Liu, B.A. Assaf, D. Gall, R. Sundararaman, and C.L. Hinkle, “PtCoO 2 for  Scaled Interconnects,” Small Struct. 2400638, (2025).  10 T. Harada, Z.P.L. Ang, Y. Sakakibara, T. Nagai, and Y. Masahiro, “Scalable  alloy-based sputtering of high-conductivity PdCoO2 for advanced interconnects,”  Appl. Phys. Lett. 126, 221902 (2025).  11 J.M. Allred, S. Jia, M. Bremholm, B.C. Chan, and R.J. Cava, “Ordered CoSn- type ternary phases in Co 3Sn 3-xGe x,” J. Alloys Compd. 539, 137–143 (2012).  12 M. Kakihana, K. Nishimura, D. Aoki, A. Nakamura, M. Nakashima, Y. Amako,  T. Takeuchi, T. Kida, T. Tahara, M. Hagiwara, H. Harima, M. Hedo, T. Nakama,  and Y. Ōnuki, “Electronic states of antiferromagnet FeSn and Pauli paramagnet  cosn,” J. Physical Soc. Japan 88, 014705 (2019).  13 W.R. Meier, M.-H. Du, S. Okamoto, N. Mohanta, A.F. May, M.A. McGuire,  C.A. Bridges, G.D. Samolyuk, and B.C. Sales, “Flat bands in the CoSn-type  compounds,” Phys. Rev. B 102(7), 075148 (2020).  14 B.C. Sales, W.R. Meier, A.F. May, J. Xing, J.-Q. Yan, S. Gao, Y.H. Liu, M.B.  Stone, A.D. Christianson, Q. Zhang, and M.A. McGuire, “Tuning the flat bands  of the kagome metal CoSn with Fe, In, or Ni doping,” Phys. Rev. Mater. 5(4),  044202 (2021).  15 H. Huang, L. Zheng, Z. Lin, X. Guo, S. Wang, S. Zhang, C. Zhang, Z. Sun, Z.  Wang, H. Weng, L. Li, T. Wu, X. Chen, and C. Zeng, “Flat-Band-Induced  Anomalous Anisotropic Charge Transport and Orbital Magnetism in Kagome  Metal CoSn,” Phys. Rev. Lett. 128(9), 096601 (2022).  16 T. Nakatani, N. Suwannaharn, and T.T. Sasaki, “Fabrication and  characterization of single-crystalline CoSn(101¯0) kagome metal thin films for  interconnect applications: Structure and anisotropic electrical resistivity,” Appl.  Phys. Lett. 127(20), 203302 (2025).  17 I. Galanakis, N. Papanikolaou, and P.H. Dederichs, “Applicability of the  broken-bond rule to the surface energy of the fcc metals,” Surf. Sci. 511(1–3), 1– 12 (2002).  18 J. Talvacchio, G.R. Wagner, and H.C. Pohl, “YBa2Cu3O7 films grown on  epitaxial MgO buffer layers on sapphire,” Physica C: Superconductivity and Its  Applications 162–164, 659–660 (1989).  19 P.A. Stampe, M. Bullock, W.P. Tucker, and R.J. Kennedy, “Growth of MgO  thin films on M-, A-, C- and R-plane sapphire by laser ablation,” J. Phys. D Appl.  Phys. 32(15), 1778–1787 (1999). 29  &#32; 20 G. Kresse, and J. Hafner, “Ab initio molecular dynamics for liquid metals,”  Phys. Rev. B 47(1), 558 (1993).  21 G. Kresse, and J. Furthmüller, “Efficiency of ab-initio total energy calculations  for metals and semiconductors using a plane-wave basis set,” Comput. Mater.  Sci. 6(1), 15–50 (1996).  22 J.P. Perdew, K. Burke, and M. Ernzerhof, “Generalized Gradient  Approximation Made Simple,” Phys. Rev. Lett. 77(18), 3865 (1996).  23 P.E. Blöchl, “Projector augmented-wave method,” Phys. Rev. B 50(24), 17953  (1994).  24 H.J. Monkhorst, and J.D. Pack, “Special points for Brillouin-zone  integrations,” Phys. Rev. B 13(12), 5188 (1976).  25 A.K. Larsson, M. Haeberlein, S. Lidin, and U. Schwarz, “Single crystal  structure refinement and high-pressure properties of CoSn,” J. Alloys Compd.  240(1–2), 79–84 (1996).  26 L. Bengtsson, “Dipole correction for surface supercell calculations,” Phys.  Rev. B 59(19), 12301 (1999).  27 F. Yalcin, and M. Wolloch, “SurfFlow: High-throughput surface energy  calculations for arbitrary crystals,” Comput. Mater. Sci. 234, 112799 (2024).  28 J.F. Nye, Physical Properties of Crystals, Their Representation by Tensors and  Matrices (Oxford University Press, 1985).  &#32;