Local structure of Nb in superconducting Nb-doped ${\mathrm{Bi}}_2{\mathrm{Se}}_3$

In the prospect of realizing bulk superconductivity in a topological insulator, metal-doped ${\mathrm{Bi}}_2{\mathrm{Se}}_3$ has been investigated with increased interest, where the Cu-, Sr-, and Nb-doped systems appear particularly promising. It is generally assumed that metal intercalation into the van der Waals (vdW) gap is responsible for the superconductivity. We have investigated the local structure of Nb in samples with nominal composition ${\mathrm{Nb}}_{0.25}{\mathrm{Bi}}_2{\mathrm{Se}}_3$ and ${\mathrm{Nb}}_{0.25}{\mathrm{Bi}}_{1.75}{\mathrm{Se}}_3$ using the x-ray absorption fine structure technique. It is found that that Nb is primarily located in a local environment consistent with that of the misfit layered structure ${\left(\mathrm{BiSe}\right)}_{1+\delta }{\mathrm{NbSe}}_2$, which has a $\delta$-dependent superconducting transition in the same temperature range. We explore the possibility of Nb occupancy on various sites in the ${\mathrm{Bi}}_2{\mathrm{Se}}_3$ structure, but neither intercalation nor substitution lead to physically meaningful improvements of the models. Furthermore, we report single crystal x-ray diffraction analysis of Nb-doped ${\mathrm{Bi}}_2{\mathrm{Se}}_3$. Difference density maps are found to show negligible occupancy in the vdW gap. The misfit layer compound has recently been suggested as an alternative origin for superconductivity in the Nb-doped ${\mathrm{Bi}}_2{\mathrm{Se}}_3$ system, in good agreement with the present study. Our findings stress the necessity of thorough structural characterization of these samples. In more general terms, it raises the question of whether metal intercalation is responsible for the superconductivity in the Cu- and Sr-doped ${\mathrm{Bi}}_2{\mathrm{Se}}_3$ systems or phase segregation plays a role as well.

Figure 1

The DFT relaxed ${\mathrm{Bi}}_2{\mathrm{Se}}_3$ structures with Nb (a) in interstitial site 1, (b) in interstitial site 2, and (c) substituting for Bi. (d) ${\mathrm{NbSe}}_2$ from Ref. [39]. (e) Schematic of the misfit structure, ${\left(\mathrm{BiSe}\right)}_{1+\delta }{\mathrm{NbSe}}_2$ based on the analogous ${\left(\mathrm{BiSe}\right)}_{1.1}{\mathrm{TaSe}}_2$ from Ref. [40]. All structures are viewed along their $a$ axis.

Figure 2

(a) Powder x-ray diffractograms of the samples S1, S2 and S3 and of synthesized reference materials. In (b), a zoom is shown at the position of the (002) peak of the misfit structure. (c) Magnetic susceptibility of the three samples measured at temperatures down to 1.8 K, at a constant field of 20 Oe. (d) Resistance of S1 and S2 measured down to 1.4 K.

Figure 3

(a) Image of the S2-crystal for the SCXRD experiment. Only the bottom part, indicated by the arrow, was probed by the beam. (b) Fluorescence spectrum for the whole S2-crystal at 18936 eV (below the Nb K-edge) and at 19036 eV (above the Nb K-edge). (c) ${\mathrm{Bi}}_2{\mathrm{Se}}_3$ unit cell showing the (110) plane. (d) The Fourier difference electron density of the (110) plane when the SCXRD data were fitted with a ${\mathrm{Bi}}_2{\mathrm{Se}}_3$ model. Red (blue) indicates excess (deficient) electron density in the data compared with the model. The red arrow shows the 3b site in the vdW gap, the black arrow shows the Bi site, and the two Se sites are indicated by blue arrows.

Figure 4

XANES spectra of the Nb K-edge measured at 300 K at (a) DND-CAT and (b) BALDER. (c, d) Linear combination fit of the misfit and ${\mathrm{Nb}}_2{\mathrm{O}}_5$ reference spectra to the S1 (c) and S2 (d) spectrum measured at DND-CAT. (e) Linear combination fit of the misfit and ${\mathrm{Nb}}_2{\mathrm{O}}_5$ reference spectra to the S3 spectrum measured at BALDER. The fits resulted in the compositions given in the graphs.

Figure 5

EXAFS data from (a, b) DND-CAT and (c, d) BALDER plotted as the $k^2$-weighted $\chi \left(k\right)$ and the magnitude of its Fourier transform (FT) $\left|\chi \right(R\left)\right|$. The FT window 3.15–15 ${\AA }^{-1}$ was used for all spectra in (b), and a window of 3.2–13.0 ${\AA }^{-1}$ was used for S3, the misfit phase reference and ${\mathrm{NbSe}}_2$ in (d). A window of 2.9–15.2 ${\AA }^{-1}$ was used for the ${\mathrm{NbO}}_2$ and ${\mathrm{Nb}}_2{\mathrm{O}}_5$. The labels in (b) and (d) are the same for graphs of the same color in (a) and (c), respectively. For S1 and S2, the solid line indicates the low temperature, and the dashed line indicates the 300 K data.

Figure 6

(a, b) The ${\mathrm{NbSe}}_2$ (representing the misfit layer compound) model fitted to the EXAFS data of (a) S1 plotted with an upward shift and S2, and (b) S3. The amplitude reduction factor was either fixed to $S_0^2=1.0$ (blue line) or a fitted variable (red line). (c, d) The magnitude of $\chi \left(R\right)$ of the DFT relaxed structures in Figs. 1–1 along with the measured EXAFS for S2 (c) and S3 (d). The DFT models were not fitted to the data but plotted with the same user-defined ${\sigma }^2$ for all single-scattering paths in all models. A FT window of 3.15–15.0 ${\AA }^{-1}$ was used in (a, c) and 3.2-13.0 ${\AA }^{-1}$ in (b, d).