Main

The search for exotic quantum phenomena and new functionalities has been among the most tremendous driving forces for the fields of condensed-matter physics and materials science. Majorana zero modes, that is, Majorana fermions that are their own antiparticles and occur at exactly zero energy, are particularly fascinating not only because of their intriguing physics obeying robust non-Abelian statistics, but also owing to their potential application as building blocks for topological quantum computers7,8. Although significant progress has been made recently in one-dimensional semiconductor quantum wires coupled with conventional superconductors9,10,11,12, decisive evidence of Majorana zero modes has been lacking and many puzzles remain13. Topological insulators, whose hallmark is time-reversal-symmetry-protected surface states, may offer less restrictive experimental conditions for realizing Majorana zero modes1,2. Theoretically, Majorana zero modes are predicted to occur in vortex cores of three-dimensional topological insulators when they are in close proximity to conventional s-wave superconductors6; however, identifying them experimentally has been challenging mainly because of the small pairing gap Δ∼1 meV induced by typical conventional superconductors14 and the extremely small energy splitting ∼Δ2/εF∼10−3 meV between Majorana zero modes and other low-lying vortex core bound states15, where εF∼400 meV is the typical Fermi energy of surface states14. A heterostructure between a topological insulator and a cuprate high-temperature superconductor (Fig. 1a), which has an order of magnitude enhancement in critical temperature Tc and Δ compared with typical conventional s-wave superconductors, may offer a more feasible approach16,17.

Figure 1: Schematic and characterization of the Bi2Se3/Bi2212 heterostructure.
figure 1

a, Schematic of the Bi2Se3/Bi2212 heterostructure and proximity-induced gap on the surface states. b, STM topography (sample bias voltage 0.3 V, tunnelling current 0.03 nA) of the BiO surface. c, Typical dI/dV spectrum showing a superconducting gap of 45 meV at 4.2 K on the surface of Bi2212. d, STM topography (100 nm×100 nm, 1.0 V, 0.03 nA) of 7 QL Bi2Se3 films grown on Bi2212. e, Line profile along the white line in d. f, Atomic-resolved STM topography (0.4 mV, 0.07 nA) shows two predominant crystal orientations of Bi2Se3.

Full size image

To harbour Majorana zero modes with robust non-Abelian statistics, nodeless superconductivity needs to be induced on the surface states of topological insulators. Although previous tunnelling junction experiments have suggested a proximity-induced gap of 10 meV at the interface between exfoliated bulk Bi2Se3 and Bi2Ca2Cu2O8+δ (Bi2212) samples18, two critical questions remain unanswered for the proximity-induced superconductivity: whether the superconducting gap is induced on the surface states or on the bulk states, and what the pairing symmetry is. To address these issues and achieve high- Tc superconductivity in topological insulators, we prepare high-quality Bi2Se3 thin films directly on a freshly cleaved Bi2212 surface by molecular beam epitaxy (MBE). MBE growth can provide high-quality topological insulator films and controllable interfaces. More importantly, MBE films are suitable for ultrahigh vacuum measurements, such as scanning tunnelling microscopy (STM) and angle-resolved photoemission spectroscopy19 (ARPES). The latter can be conveniently used to determine whether the gap is on the topological surface states or on the bulk bands, measure the gap size and reveal the pairing symmetry. Our ARPES measurements reveal a proximity-induced gap up to 15 meV, and this apparent gap is located only on the topological surface states with predominant s-wave pairing. One possible explanation for this unusual proximity effect is the combined consequence of disorder and the relative lattice orientation between cuprate and topological insulator film. Our work not only provides new opportunities for investigating the intriguing coupling between a topological insulator thin film and a d-wave superconductor, but also greatly expands the experimental range for realizing robust Majorana zero modes and achieving new functionalities.

Figure 1a shows a schematic of the Bi2Se3/Bi2212 heterostructure. STM topography (Fig. 1b) and scanning tunnelling spectroscopy (Fig. 1c) confirm the high surface quality and superconducting property of the cleaved Bi2212 sample. The growth of Bi2Se3 thin films on Bi2212 is exceptionally challenging for two reasons: the low surface binding energy of Bi2212 prefers the island growth mode to the layer-by-layer mode; oxygen dopants in Bi2212 can easily oxidize the films grown on it when heated above 200 °C, a typical temperature for growing Bi2Se3. By carefully controlling the growth kinetics, we have succeeded in growing high-quality Bi2Se3/Bi2212 heterostructures in the layer-by-layer growth mode so that high-resolution STM and ARPES measurements can be carried out. Figure 1d–f shows the high quality of the 7 quintuple layer (QL) thick Bi2Se3 film on Bi2212. Although Bi2Se3 and Bi2212 have very different (hexagonal versus tetragonal) crystal structures, the Bi2Se3 films grown on Bi2212 show only two predominant crystal orientations with Se–Se bonds parallel to the x and y (anti-nodal, nearest Bi–Bi bond) directions of Bi2212, respectively (Fig. 1b,f and Supplementary Information).

The existence of topological surface states in the Bi2Se3/Bi2212 heterostructure can be revealed in the ARPES data measured at 50 eV photon energy shown in Fig. 2. Near the Fermi energy (EF), the intensity map of the Dirac cone shows a circular shape, suggesting that the electronic structure is rather isotropic. The electron pocket decreases in size with binding energy and merges into a single point at −0.4 eV (Fig. 2a). Figure 2c shows a cut through the Γ point. Here the conical dispersion is observed from EF to −0.6 eV, where the bulk valence bands start to arise. The electronic structure is in good agreement with the conical dispersion reported on the surface states in both bulk and thin film Bi2Se3 samples14,20,21,22. The shift of chemical potential to 0.4 eV above the Dirac point is possibly attributed to charge transfer from the substrate and to defects in the Bi2Se3 film. In addition to surface states and valence bands, bulk conduction bands, shown as quantum well-like states (QWSs) near the Fermi energy, are also expected in thin-film Bi2Se3 samples14,22. However, the surface states and QWSs have different dipole matrix elements23, which makes it possible to separate their contributions by choosing an appropriate photon energy (Supplementary Information for detailed photon energy dependence). At 50 eV photon energy, the QWSs are strongly suppressed, thus allowing us to resolve the surface states without being affected by the bulk bands.

Figure 2: ARPES data measured at 20 K with 50 eV photon energy reveal the nearly isotropic gap on the topological surface states.
figure 2

a, Constant energy maps. b, MDC integrated from −20 to 20 meV. c, ARPES data. Dotted lines are guides for the dispersions of surface states (SS) and valence band. d, EDCs at momentum positions marked in c. Red tick marks are guides for the peak positions of the surface states. e,f, Zoom-in of EDCs at kL (e) and kR (f). Black lines are reference spectra for EF. g, Azimuth (ϕ) dependence of the gap size (radial axis) along the Fermi surface.

Full size image

We analyse the energy distribution curves (EDCs) taken at a photon energy of 50 eV (Fig. 2d) to detect the proximity-induced gap on the surface states of Bi2Se3. The linear dispersions of the surface states can be observed in the EDCs, with peaks approaching the Fermi energy at two Fermi momenta kL and kR. We further zoom-in the EDCs at these two momentum positions (Fig. 2e,f). A comparison of the EDCs with a reference spectrum taken from a clean metal surface shows clearly that there is a suppression of intensity near EF, that is, a gap is open. Following the standard ARPES procedure to extract the gap by the shift of leading edge position24, we obtain the gap size of 15 meV on the topologically protected surface states. We note that although this large gap can be observed only on films with extremely high quality (sharper ARPES peaks), a smaller gap of 6–8 meV is routinely observed in regular samples measured so far, which is still an order of magnitude larger than that induced by a conventional s-wave superconductor14. Surprisingly, despite the d-wave pairing of Bi2212 that is strongly anisotropic with four nodes of zero gap24, the induced gap on the surface states is rather isotropic (Fig. 2g) and consistent with s-wave pairing.

We examine the data taken at 20 K with photon energy of 30 eV where there is a strong intensity contribution from QWSs to determine whether a gap is induced on the bulk conduction bands. The strong intensity contribution from QWSs is evident in Fig. 3a–c. Here the lower QWS (labelled as QWS2) is clearly resolved to overlap with the surface states between EF and −200 meV with a nearly flat band bottom at −200 meV. The inner QWS (labelled as QWS1) shows up as a smaller pocket near EF. The EDCs at the Fermi momenta for QWS2 (pink and red curves) and QWS1 (orange and green curves) shown in Fig. 3d,e set an upper limit of 3 meV on the gap size for these QWSs. Data taken with other photon energies, for example 48 and 52 eV, where there are contributions from both the surface states and the QWSs, show that the gap size increases consistently with increasing contribution from the surface states (Supplementary Information). This observation further supports that the large gap does reside on the surface states and that the gap on the QWSs is significantly smaller. We note that the gap on the surface states is unlikely to be detected directly by STM measurements, because STM lacks the capability to distinguish surface states from bulk QWSs. Moreover, STM signals are dominated by QWSs that have negligibly small gap and yet have much higher density of states near the Fermi energy (for example, see the flatter band from QWS1 near EF).

Figure 3: ARPES data measured at 20 K with 30 eV photon energy show a much smaller gap on the QWSs.
figure 3

a, MDC integrated from −20 to 20 meV. b, ARPES data. The colour scale is the same as in Fig. 2a. c, EDCs at momentum positions marked in b. Black tick marks are guides for the peak positions of the QWSs. d,e, Zoom-in of EDCs at Fermi momenta (coloured curves in c) near EF. Here the gap on the QWSs is reduced to less than 3 meV.

Full size image

The temperature-dependent ARPES data measured with 50 eV photon energy further reveal that the gap on the surface states is induced by superconductivity in Bi2212. Figure 4 shows EDCs at Fermi momenta measured at various temperatures on a sample with 7 meV gap. The gap is observed at least up to 50 K. Above 50 K, the gap becomes much harder to detect and completely vanishes at 100 K. To check reproducibility, we have subsequently cooled down the sample from 110 K directly to 20 K and confirmed that the gap at 20 K was reproducible (bottom curves in Fig. 4a,b). Figure 4c shows the extracted gap size as a function of temperature. The similar temperature-dependent behaviour as the superconductivity transition for Bi2212 clearly suggests that the gap observed on the surface states of Bi2Se3 is closely related to superconductivity in Bi2212 through the proximity effect.

Figure 4: Temperature dependence of the gap on the surface states measured with 50 eV photon energy.
figure 4

a,b, EDCs at kL and kR (labelled in Fig. 2b) measured at different sample temperatures. Black curves were reference spectra for EF. c, Extracted gap as a function of temperature. The gap values are normalized by the average gap size at 20 K. The gap is defined when the shifted EDC at kL (kR) shows best overlap with the EF reference spectrum, and the error bar is defined when the shifted EDCs starts to show clear deviation from the EF reference spectrum.

Full size image

The coupling between a d-wave superconductor and a topological insulator is an intriguing question. The observation of a nearly isotropic superconducting gap on the topological surface states and a much smaller gap on the bulk conduction bands is a reflection of such non-trivial coupling. If the crystalline group of the Bi2Se3 film were tetragonal like Bi2212, the proximity-induced pairing gap in the Bi2Se3 film would be expected to be nodal d-wave in the clean limit. However, because the Bi2Se3 film possesses hexagonal crystal symmetry, the proximity-induced paring gap in the Bi2Se3 is in principle a mixture of s-wave and d-wave. The s-wave component of the induced pairing in the Bi2Se3 is zero when the nodal direction of Bi2212 lies in one of the reflection planes of the Bi2Se3 for the following symmetry reasons. If this occurs, the hybrid Bi2Se3/Bi2212 system would obey the reflection symmetry along the nodal direction. Under such reflection, the s-wave pairing in the Bi2Se3 is even whereas the d-wave pairing in Bi2212 is odd. Fortunately, it is not the case for the Bi2Se3/Bi2212 heterostructure here (Fig. 1b,f). Indeed, it turns out that the Bi2Se3 reflection directions are at 15° or 45° from cuprate nodal directions (Supplementary Information); consequently, a sizable s-wave component is expected in the induced pairing of the topological insulator. Furthermore, because of the smaller gap and longer coherence length ξ≈2ℏvF/(πΔ)≈14.4 nm in Bi2Se3 compared with cuprates (≈1 nm), the d-wave gap is much more fragile against disorder and the formation of small domains; however, the s-wave gap is robust by Anderson’s theorem25. Experimentally, we have observed that the pairing gap on the topological insulator’s surface states is relatively large and nearly isotropic in the entire momentum space (Fig. 2g) although the pairing gap on the QWSs is negligible. A plausible explanation for this distinction is that the topological insulator’s surface states have a predominant s-wave pairing gap that is robust against disorder as well as the formation of domains, whereas the bulk QWSs have a predominant d-wave pairing gap that is significantly suppressed by disorder. If the gap on the topological surface states of Bi2Se3 is confirmed to be s-wave, this will open up a large range of experimental opportunities for observing and engineering a single Majorana zero mode. For instance, STM measurements might be able to distinguish this single Majorana zero mode in a vortex core from other low-lying quasi-particle bound states because the energy splitting between them is of the order of Δ2/εF∼0.5 meV, which is within the energy resolution of the state-of-art STM. We note that non-Abelian statistics of the putative Majorana zero mode in a vortex core may not be sufficiently robust owing to the QWSs with vanishingly small gap. Tuning the chemical potential away from QWSs should avoid such complication from QWSs.

Methods

The MBE growth and in situ STM experiments were conducted in an MBE–STM combined system with a base pressure of 1.0×10−10 Torr. Single-crystal Bi2212 samples with optimum oxygen doping (Tc≈91 K) were grown by the travelling floating zone method26. The Bi2212 samples were cleaved in ultrahigh vacuum at room temperature, and annealed at 250 °C for 3 h to degas the sample holder to avoid contamination of the film caused by outgassing of the sample holder during the growth process. The atomic structures of Bi atoms on the BiO-terminated surface of Bi2212 and the well-known supermodulation in the STM topographic image27 (Fig. 1b) reveal the high quality of the cleaved surface. The characteristic V-shaped gap of 45 meV in the dI/dV spectrum in Fig. 1c confirms that the Bi2212 is still superconducting at near optimum doping and there is only a slight change in the oxygen doping after annealing (Supplementary Fig. S1).

Bi2Se3 films were then grown on Bi2212 at 190 °C by co-evaporating Bi (99.9999%) and Se (99.999%) from standard Knudsen cells under a Se-rich condition with a growth rate of approximately 0.07 QL per minute22. STM images were taken at 4.2 K with a polycrystalline PtIr tip. The STM image in Fig. 1d shows an atomically flat Bi2Se3 film on Bi2212 with a nominal thickness of 7 QL. Small regions of 6 QL are also present as indicated by the line profile in Fig. 1e. The step height of 0.95 nm is in agreement with the thickness of 1 QL Bi2Se3 (ref. 28). The atomically resolved STM image (Fig. 1f) reveals a hexagonal lattice structure with a lattice constant of ≈0.4 nm, consistent with the Se-terminated (111) surface of Bi2Se3 (ref. 28).

ARPES measurements were taken at BL12.0.1 of the Advanced Light Source at Lawrence Berkeley National Laboratory. Before being transferred from the STM chamber to the ARPES chamber, the samples were covered with an amorphous Se protection capping layer, which was removed in the ARPES chamber by annealing at 200 °C for 30 min until sharp core levels (Supplementary Information) and clear dispersions were observed. The Fermi energy was calibrated by reference spectra taken from a clean metal (gold or tantalum) surface.