Highly Efficient Spin-to-Charge Current Conversion in Strained HgTe Surface States Protected by a HgCdTe Layer

We report the observation of spin-to-charge current conversion in strained mercury telluride at room temperature, using spin pumping experiments. We show that a HgCdTe barrier can be used to protect the HgTe from direct contact with the ferromagnet, leading to very high conversion rates, with inverse Edelstein lengths up to $2.0\pm 0.5\mathrm{nm}$. The influence of the HgTe layer thickness on the conversion efficiency is found to differ strongly from what is expected in spin Hall effect systems. These measurements, associated with the temperature dependence of the resistivity, suggest that these high conversion rates are due to the spin momentum locking property of HgTe surface states.

Figure 1

(a) Schematic representation of the band structure of strained HgTe, with the Dirac dispersion cone of the surface states, and the bulk ${\mathrm{\Gamma }}_8$ band. The arrows represent the helical spin configuration. (b) X-ray reflectivity spectrum of a HgTe $(18.5\mathrm{nm})/\mathrm{HgCdTe}$ (5.5 nm) sample. The structure used for the fit is represented in the inset. The red dashed curve represents the experimental data; the black curve is the fit. (c) Scanning tunneling electron microscopy HAADF image of $\mathrm{HgCdTe}/\mathrm{HgTe}/\mathrm{HgCdTe}$ structure and corresponding chemical profile.

Figure 2

(a) Principle of the spin pumping FMR measurement. (b) Broadband frequency dependence of the peak-to-peak FMR linewidth of the reference $\mathrm{NiFe}/\mathrm{Si}$ sample and of a $\text{NiFe}/\text{HgCdTe}(1.6)/\text{HgTe}(18.5)$ sample. The damping coefficient of NiFe is higher when deposited on HgTe (${\alpha }_{\mathrm{ref}}=6.33\times {10}^{-3}\pm 3\times {10}^{-5}$ for the reference sample compared to $\alpha =7.50\times {10}^{-3}\pm 7\times {10}^{-5}$). For a rf field of 0.1 mT, this leads to a pure spin current $J_s^{3\mathrm{D}}=7.6\pm 0.2\mathrm{MA}/{\mathrm{m}}^2$. (c) FMR and dc voltages, measured by spin pumping FMR on the same sample. The symmetric (red) and antisymmetric (green) contributions have been extracted from the measured signal (in blue). (d) Spin pumping signals obtained for a positive and a negative dc field, on the same sample.

Figure 3

(a) Spin pumping signals obtained for different thicknesses of HgCdTe barriers, normalized by the sample resistance. (b) HgCdTe thickness dependence of $I_C/w$. The measurements were all performed on HgTe layers of the same thickness (18.5 nm), while varying the HgCdTe layer thickness using Ar etching. The thicknesses were measured by XRR. (Inset) Scheme of the conversion. (c) Sheet resistance $R_{\text{sheet}}$ as a function of the temperature, for three samples of different HgTe thicknesses (8.5, 18.5, and 56 nm). (d) HgTe thickness dependence of the inverse Edelstein length ${\lambda }_{\mathrm{IEE}}$. The HgCdTe barrier thickness is the same for all the samples ($t_{\mathrm{HgCdTe}}\sim 1.6\mathrm{nm}$). The large error bars for the 26 and 84 nm thick samples are due to a relatively large uncertainty on the extra damping.