Superconducting proximity effect in a topological insulator using Fe(Te, Se)

Interest in the superconducting proximity effect has recently been reignited by theoretical predictions that it could be used to achieve topological superconductivity. Low-$T_c$ superconductors have predominantly been used in this effort, but small energy scales of $\sim 1$ meV have hindered the characterization of the emergent electronic phase, limiting it to extremely low temperatures. In this work, we use molecular beam epitaxy to grow topological insulator $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ in a range of thicknesses on top of a high-$T_c$ superconductor Fe(Te,Se). Using scanning tunneling microscopy and spectroscopy, we detect ${\mathrm{\Delta }}_{\mathrm{ind}}$ as high as $\sim 3.5$ meV, which is the largest reported gap induced by proximity to an $s$-wave superconductor to date. We find that ${\mathrm{\Delta }}_{\mathrm{ind}}$ decays with $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ thickness, but remains finite even after the topological surface states have been formed. Finally, by imaging the scattering and interference of surface state electrons, we provide a microscopic visualization of the fully gapped $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ surface state due to Cooper pairing. Our results establish Fe-based high-$T_c$ superconductors as a promising new platform for realizing high-$T_c$ topological superconductivity.

Figure 1

(a) Schematic of $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$/Fe(Te,Se) heterostructure. STM measurements are probing the exposed top surface of $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$. (b) Typical postgrowth RHEED image of the MBE-grown heterostructure showing long streaks characteristic of two-dimensional layer-by-layer growth. (c) STM topograph showing (d) the exposed Fe(Te,Se) substrate with square atomic lattice and (e) the $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ film with hexagonal lattice structure. (f) Height profile taken along the dashed red line in (c) consistent with 1 QL of $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ grown on top of Fe(Te,Se). STM setup condition in: (c) $I_{\mathrm{set}}=10\mathrm{pA}$, $V_{\mathrm{set}}=\text{–}200\mathrm{mV}$; (d) $I_{\mathrm{set}}=30\mathrm{pA}$, $V_{\mathrm{set}}=\text{–}50\mathrm{mV}$; (e) $I_{\mathrm{set}}=30\mathrm{pA}$, $V_{\mathrm{set}}=\text{–}50\mathrm{mV}$.

Figure 2

Thickness-dependent quasiparticle interference imaging of the $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ surface states. (a), (e), (i) Typical STM topographs, (b), (f), (j) large region $dI/dV$ maps acquired at 50 mV, (c), (g), (k) their associated Fourier transforms and (d), (h), (l) extracted QPI dispersion along Γ-M direction for 1 QL, 3 QL, and 5 QL thick $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ films, respectively. Thick gray lines in (d), (h), (l) represent fit to our data based on ARPES measurements in Ref. [22], shifted upwards by (d) $\sim 150$ meV, (h) $\sim 100$ meV, and (l) $\sim 100$ meV to match the QPI dispersions. This rigid band shift can be easily explained by the difference in substrate choice and possibly different impurity concentration, as the shift of the same magnitude is also observed between ARPES studies of for example 1 QL $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3/\mathrm{NbS}{\mathrm{e}}_2$ [9] and 1QL $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3/\mathrm{Si}$ [22]. The FTs in (g), (k) have been sixfold symmetrized to increase the signal-to-noise. STM setup condition in: (a) $I_{\mathrm{set}}=80\mathrm{pA}$, $V_{\mathrm{set}}=\text{–}40\mathrm{mV}$; (b) $I_{\mathrm{set}}=450\mathrm{pA}$, $V_{\mathrm{set}}=\text{–}150\mathrm{mV}$, $V_{\mathrm{exc}}=10\mathrm{mV}$ (zero-to-peak); (e) $I_{\mathrm{set}}=30\mathrm{pA}$, $V_{\mathrm{set}}=6\mathrm{mV}$; (f) $I_{\mathrm{set}}=100\mathrm{pA}$, $V_{\mathrm{set}}=50\mathrm{mV}$, $V_{\mathrm{exc}}=10\mathrm{mV}$; (i) $I_{\mathrm{set}}=100\mathrm{pA}$, $V_{\mathrm{set}}=10\mathrm{mV}$; (j) $I_{\mathrm{set}}=50\mathrm{pA}$, $V_{\mathrm{set}}=50\mathrm{mV}$, $V_{\mathrm{exc}}=10\mathrm{mV}$.

Figure 3

Temperature dependence of $dI/dV$ spectra and Abrikosov vortex lattice imaging. (a) Average $dI/dV$ spectrum on 1 QL $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ as a function of temperature, vertically offset for clarity, showing the gap closing at $\sim 10\mathrm{K}$. All spectra have been acquired under the same conditions over the identical 20 nm region of the sample. (b)–(g) $dI/dV$ maps acquired at −3 mV bias over the region of the sample shown in Fig. 2 in a range of magnetic fields from 0.15–9 T. Each dark region in (b)–(g) represents a single vortex, and the vortex lattice is visible up to the highest field studied. (h) Radially averaged $dI/dV$ spectra as a function of distance from the center of the vortex [0.15 T in (b)]. Zero bias peak and the suppression of coherence peaks are apparent in the vortex center (darkest blue curve). STM setup condition in: (a) $I_{\mathrm{set}}=60\mathrm{pA}$, $V_{\mathrm{set}}=\text{–}10\mathrm{mV}$, $V_{\mathrm{exc}}=0.2\mathrm{mV}$; (b)–(g) $I_{\mathrm{set}}=3\mathrm{pA}$, $V_{\mathrm{set}}=\text{–}3\mathrm{mV}$, $V_{\mathrm{exc}}=1.5\mathrm{mV}$ (zero-to-peak); (h) $I_{\mathrm{set}}=60\mathrm{pA}$, $V_{\mathrm{set}}=\text{–}10\mathrm{mV}$, $V_{\mathrm{exc}}=0.45\mathrm{mV}$.

Figure 4

Evolution of the induced superconducting gap ${\mathrm{\Delta }}_{\mathrm{ind}}$ with $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ film thickness. (a) Average $dI/dV$ spectra obtained on: Fe(Te,Se) substrate (black), 1 QL (red), 3 QL (blue), and 5 QL (green) $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ film grown on superconducting Fe(Te,Se). (b) ${\mathrm{\Delta }}_{\mathrm{ind}}$ as a function of $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ film thickness. Dashed line in (b) represents the decay rate of the ${\mathrm{\Delta }}_{\mathrm{ind}}$ on the surface of $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ induced by $\mathrm{NbS}{\mathrm{e}}_2$ from Ref. [8], scaled by a multiplicative factor to match the ${\mathrm{\Delta }}_{\mathrm{ind}}$ for 1 QL in our case. STM setup condition in: (a) (black) $I_{\mathrm{set}}=30\mathrm{pA}$, $V_{\mathrm{set}}=10\mathrm{mV}$, $V_{\mathrm{exc}}=0.5\mathrm{mV}$ (zero-to-peak); (red) same as Fig. 3; (blue) $I_{\mathrm{set}}=30\mathrm{pA}$, $V_{\mathrm{set}}=5\mathrm{mV}$, $V_{\mathrm{exc}}=0.4\mathrm{mV}$ (zero-to-peak); (green) $I_{\mathrm{set}}=30\mathrm{pA}$, $V_{\mathrm{set}}=6\mathrm{mV}$, $V_{\mathrm{exc}}=0.4\mathrm{mV}$ (zero-to-peak).

Figure 5

Suppression of QPI within the superconducting gap. (a)–(i) Fourier transforms of $dI/dV$ maps acquired over the region in Fig. 2. While distinct quasi-circular QPI peak can be seen in (a)–(d), the same peak is obviously suppressed in (e)–(i). Normalized QPI intensity relative to the background as a function of STM bias for (j) 1 QL thick and (k) 3 QL thick $\mathrm{B}{\mathrm{i}}_2\mathrm{T}{\mathrm{e}}_3$ films [34]. Position of the orange lines in (j), (k) indicate the approximate energy of the observed gap peaks, while the width of the lines corresponds to thermal broadening expected at the measurement temperature. All data in (a)–(i) have been acquired under the same experimental conditions (0.33 GΩ junction resistance with 1 mV zero-to-peak lock-in excitation).