Metallic phase in stoichiometric CeOBiS2 revealed by space-resolved ARPES.

Recently CeOBiS2 system without any fluorine doping is found to show superconductivity posing question on its origin. Using space resolved ARPES we have found a metallic phase embedded in the morphological defects and at the sample edges of stoichiometric CeOBiS2. While bulk of the sample is semiconducting, the embedded metallic phase is characterized by the usual electron pocket at X point, similar to the Fermi surface of doped BiS2-based superconductors. Typical size of the observed metallic domain is larger than the superconducting correlation length of the system suggesting that the observed superconductivity in undoped CeOBiS2 might be due to this embedded metallic phase at the defects. The results also suggest a possible way to develop new systems by manipulation of the defects in these chalcogenides with structural instability.

Introduction

The discovery of superconductivity in BiS2-based materials[1] has stimulated large interest aiming at the search of superconductors with higher transition temperature in this new family of layered chalcogenides[2,3]. Beyond this, the BiS2-based systems are also found to have large potential in the field of thermoelectrics[4,5] due to highly susceptible nature of their structure that can be manipulated by external conditions including chemical and physical pressures[2,3]. The structural susceptibility is related with the defect chemistry of bismuth ion that makes BiS2 square lattice highly instable[6] with the lower energy state being a disordered state characterized by coexistence of different low symmetry structural configurations[7,8].

There are now several known BiS2-based materials with majority of them having a general formula of REOBiS2 (RE = rare earth element) in which the electronically active BiS2 layers are separated by REO spacer layers. This reflects an evident structural similarity of them to the iron-based LaOFeAs superconductors[9] containing FeAs layers separated by REO spacers. The REOBiS2 systems are band insulators and substitution of F for O in REO introduces electron-doping in the active BiS2 layers. This gives rise to an electron pocket of Bi 6p character, appearing at X point of the square Brillouin zone[10,11].

Among REOBiS2 materials, CeOBiS2 is peculiar in which Ce appears in the mixed valence state of Ce3+ and Ce4+. When doped by substitution in the REO layers, Ce(O,F)BiS2 shows coexistence of superconductiviy and magnetism at low temperature[12]. The fact that stoichiometric undoped CeOBiS2 compound manifests mixed valence[13] one may expect the extra charge in CeO-layer to dope the BiS2-layer, as the case of extrinsic doping by substitution in which extra charge is placed by F in place of O. Such a situation is known to occur in so-called “self-doped” EuFBiS2 superconductor[14] in which Eu appears in mixed valence state with coexistence of Eu2+ and Eu3+. Very recent observation of superconductivity in stoichiometric CeOBiS2 system[15] may therefore apparently support the analogy with the self-doped systems. However, the interplay between the rare-earth mixed valence and the rare-earth-to-Bi charge-transfer is not that simple. CeOBiS2 single crystals with Ce3+ and Ce4+ mixed valence as well as EuFBiS2 single crystals with Eu2+ and Eu3+ mixed valence are not superconducting although some electrons are introduced to the BiS2 layer. Indeed, angle-resolved photoemission spectroscopy (ARPES) data of non-superconducting CeOBiS2[16] and EuFBiS2[17] show absence of the Bi 6p electron pockets at the Fermi level. This discrepancy indicates that there may be some new physical mechanism active for the charge transfer between the rare-earth and Bi sites. In this context, the recent observation of superconductivity in undoped CeOBiS2[15] is highly interesting and needs further investigations.

Recently, space resolved ARPES is getting known as an important experimental tool to study inhomogeneous materials[18-21] providing wealth of information on their electronic structure. Here, to address the above question of observed superconductivity without Fermi surface, we have performed space-resolved ARPES on a single crystal of undoped stoichiometric CeOBiS2. The ARPES results obtained using submicron beam size reveal that bulk of the CeOBiS2 is electronically homogeneous and insulating without any kind of microscale texturing that may be associated with the mixed valence of Ce. Incidentally, we have found metallic phase embedded in the morphological defects and at the sample edges. This metallic phase is characterized by the usual electron Fermi surface pocket at X point, similar to the doped BiS2-based superconductors[22-25]. This unexpected result may provide a possible way to understand the observed superconductivity in undoped CeOBiS2. In addition of providing a plausible interpretation of superconductivity in undoped CeOBiS2, these results may also suggest a possible way to develop new materials by manipulation of defects in instable structures.

Figure 1 shows scanning photoelectron microscopy (SPEM) maps measured on CeOBiS2 sample at 50 K. Fig. 1(a) and (b) represent respectively maps obtained by integrating photoemission intensities within eV and eV where E − E represents energy relative to the Fermi level (E). Apparently, Fig. 1(a) and (b) reveal that majority of the sample is electronically homogeneous within the spatial resolution. However, a clear contrast can be seen in the map obtained by integrating intensity in the energy range of eV, indicating that the system contains some inhomogeneity here and there. This contrast (shown by the bright and dark regions in Fig. 1(b)) is most likely due to different phases characterized by different density of states near E since this contrast appears when the integrated range is limited to eV. These differences can be better identified in the image shown in Fig. 1(c) and (d), measured with the spatial resolution of 1 × 1 μm2 (the region shown by rectangle in Fig. 1(a) or 1(b)). Figure 1(e) shows the angle- and space-integrated photoemission spectrum measured at the center of CeOBiS2 sample (majority texture). Three peak structures can be identified in the photoemission spectrum; one is located around −1.1 eV due to Ce 4f electrons[16], and the other two structures are around −2.0 eV and −2.8 eV, mainly due to S 3p contributions[10]. The integrated energy ranges are indicated in Fig. 1(e) by ‘wide’ and ‘narrow’.

Figure 1

Scanning photoelectron microscopy (SPEM) maps measured on CeOBiS2 at 50 K using hv = 27 eV. The overview SPEM image is produced by integrating photoemission intensity within the energy interval of eV (a) and eV (b). Spatial resolution for the overview SPEM image is 15 × 15 μm2. (c) and (d) are the high resolution SPEM images measured with 1 × 1 μm2 resolution (rectangular region of (a) or (b)). The rectangular region has been chosen considering a defect away from the sample edge. (e) Angle- and space-integrated photoemission spectrum. Integrated energy ranges for SPEM images are denoted by ‘wide’ and ‘narrow’ in the photoemission spectrum.

The electronic structure of Ce 4f in CeOBiS2 is similar to what has been measured earlier[16]. It is important to note that the high spectral density phase appears only around the sample edges and around morphological defects. In order to investigate the electronic structure of different phases, we have measured angle-integrated and angle-resolved photoemission (ARPES) spectroscopy in the two phases. These measurements are performed in the region ‘A’ and ‘B’ (in Fig. 1(c,d)) using sub-micron beam size. It is worth mentioning that the region ‘A’ was chosen for ARPES to avoid any possible artefact of sample edge while the integrated spectra were checked to be similar indicating that they should be from the same phase. Figure 2 shows the angle-integrated photoemission spectra measured in A and B points. The electronic structure is substantially different between the two phases. The most important difference is the structure around −0.2 eV, which appears to crosse E. This difference shows that the two phases seen in Fig. 1(b) and (d) are indeed characterized by very different spectral weight in the vicinity of E.

Figure 2

Angle-integrated photoemission spectra measured in the A- and B-points of SPEM maps of CeOBiS2 (Fig. 1(c,d)).

The next question is the nature of the two phases and if the two are characterized by some dispersive bands and Fermi surfaces. This can be clarified by the ARPES on the two phases measured in the A and B regions. Figure 3(a) and (b) are the Fermi surface maps for the two phases. The Fermi surface maps clearly show that the majority phase (region B) is non-metallic while the minority phase (region A) is metallic. Indeed, the typical Fermi surfaces of the doped BiS2-based systems[11,24,25], characterized by the electron pockets around X point, can be clearly seen in Fig. 3(a) whereas it is absent in Fig. 3(b). The ARPES on the majority phase is consistent with the earlier reports on undoped semiconducting system[16]. The presence and absence of Fermi surfaces in different regions of the sample confirm that the metallic and semiconducting phases are coexisting in stoichiometric CeOBiS2. The band dispersions along high symmetry lines of M-Γ-X-M of the Brillouin zone are shown in Fig. 3(c) and (d). As seen in the photoemission spectra (Fig. 2) and the Fermi surfaces (Fig. 3(a) and (b)), the presence/absence of the electron pockets near E is the intelligible difference between the two phases. The other features are basically the same except the spectral weight, also seen in the photoemission spectra (Fig. 2). It should be mentioned that no rigid shift has been found in photoemission studies on CeOBiS2 system as a function of charge doping induced by F-substitution in place of O[26]. Here, the average shift between different features in Figs 2 and 3 is ∼0.1–0.2 eV, consistent with earlier study on the same system.

Figure 3

Fermi surfaces for metallic phase at A-point (a) and those for semiconducting phase at B-point (b) (A- and B-points are indicated in Fig. 1(c,d)). The corresponding band dispersions along M-Γ-X-M are shown in (c) and (d), respectively.

Let us discuss briefly possible implications of the present results on the observed metallic phase and possibly the superconductivity in the stoichiometric CeOBiS2 system. The space-resolved ARPES results have clearly shown that metallic phase appears embedded in the majority texture of semiconducting phase in CeOBiS2. The electronic structure of the metallic phase is characterized by dispersing band structure and Fermi surface pocket around the X point of Bi 6p nature, typical of doped BiS2-based superconducting materials[11,24,25]. Incidentally, the metallic phase is found only around the morphological defects in the crystal while the majority of the sample is highly homogeneous and reveals usual semiconducting characteristics of BiS2-based systems without doping or self-doping[10]. Nevertheless, the specific band structure of the metallic phase indicates that this phase is not due to any extrinsic defects but it should be intrinsic to the studied sample. It is also known that, the BiS2-based systems are characterized by highly instable BiS2 square lattice[7,8] that makes the properties of these materials highly susceptible to the external conditions including chemical and physical pressures[2,3]. On the other hand, Ce in CeOBiS2 appears in mixed valence state with coexisting Ce3+ and Ce4+ and hence extra electrons are available for charge transfer from the CeO-layer to the BiS2-layer.

Here, it should be noted that the Ce mixed valence state is highly homogeneous revealed by space resolved micro X-ray absorption spectroscopy (microXAS)[16]. If the semiconducting region is similar to non-superconducting CeOBiS2 and the small metallic region is driven by electron doping due to chemical defects, the Ce valence should be different between the semiconducting and metallic regions and should exhibit inhomogeneous distribution. Therefore, the present observation suggests that the inhomogeneous electronic state of the BiS2 layer is not strictly related to the Ce valence. We think that the metallic phase should be stimulated by morphological defects due to change in the local structure around them.

It has been proposed earlier[16] that the metallic and semiconducting phases have local structure configurations depicted in Fig. 4[13] and that, in the homogeneous semiconducting region, the self-doped electrons are trapped in Bi 6p orbitals due to intrinsic local distortions[13,27-29] while in the metallic phase they remain mobile in the Bi 6p due to reduced disorder in the BiS2 square lattice. The Bi 6p electrons are randomly distributed in the lattice and do not provide a dispersive band. As pointed out earlier, the broad feature within the band gap can be assigned to the Bi 6p electrons[16,17]. It is difficult to see exact spectral weight transfer from the Bi 6p to Bi 6p since the former is broadly distributed in the momentum space. Considering all these facts it is plausible to think that the observed metallic phase around the morphological defects (including samples edges) is induced by local strain (in the instable BiS2 square lattice) and extra electrons in the CeO-layer (due to mixed valence of Ce). Here it is worth mentioning that although we have put forward a proposal based on structural instability, we are not ruling out completely any peculiar off- stoichiometry or chemical inhomogeneity to drive the metallic phase characterized by energy bands exactly similar to the doped BiS2-layer.

Figure 4

Possible local structure configurations for the semiconducting phase (a) and for the metallic phase (b) in CeOBiS2.

Therefore, one possible cause of the recently observed superconductivity in undoped stoichiometric CeOBiS2 compound could be the embedded metallic phase in the homogeneous insulating texture. A strong enough intergrain coupling can turn the system into a superconductor at low temperature as in granular superconductors[30,31]. In the studied crystal the volume fraction of the metallic phase is too small for grain coherence to induce bulk superconductivity. In this limit, an insulating behaviour is expected at low temperature at which pairs might have formed locally (locally superconducting) but the pairs remain confined inside the grains (no bulk superconductivity)[30,31]. It should be recalled that the superconductivity coherence length in these materials is less than ∼100 nm[32] and ARPES with higher space resolution may be helpful to address the exact role of electron inhomogeneity in the superconductivity of these systems.

In summary, we have performed space-resolved photoemission spectroscopy on stoichiometric CeOBiS2 system using sub-micron beam size. Using the SPEM imaging we have found a metallic granular phase embedded in the homogeneous semiconducting phase in the undoped system. The metallic phase appears around the morphological defects and is characterized by electron pockets on the Fermi surface, known for the doped BiS2-based superconducting materials. We have argued that this metallic phase is formed by the self-doping in the local symmetry broken BiS2-square lattice in the proximity of morphological defects. The Fermi surface topology is consistent with the charge-transfer from the mixed valence Ce indicating that the stoichiometric CeOBiS2 can be superconducting due to the self-doped carriers in the Bi 6p orbitals. Therefore, CeOBiS2 system, even undoped can show inhomogeneous superconductivity driven by the metallic phase embedded in the insulating texture. The present results may have direct implications on the possible way to develop new materials by manipulation of granular defects in systems with structure instability as the case of Bi-based dichalcogenides.

Methods

Sample synthesis and characterization

High-quality single crystals of stoichiometric CeOBiS2, prepared by CsCl flux method[33], were used for the space-resolved ARPES measurements. The sample used for the present work is non-superconducting down to 2 K. The sample is well characterized for its average structure and transport properties and the details are reported in ref.[33] alongwith the synthesis method.

Spectromicroscopy measurements

The experiments were carried out at the spectromicroscopy beamline of Elettra synchrotron radiation facility in Trieste, Italy[34]. Linearly polarized light of energy hv = 27 eV, focused using a Schwarzschild optics down to 500 × 500 nm2 beam spot, was falling at 45° with respect to the flat ab-plane of the single crystal sample for the present measurements. Fermi surface mapping was carried out by changing the position of electron energy analyzer with the photon beam and the sample position fixed. As for the surface treatment of the sample, we cleaved the single crystalline sample at 50 K in situ in ultrahigh vacuum (<10−10 mbar) in order to obtain a clean (001) surface. The total energy resolution including both monochromator and electron energy analyzer was measured to be ∼100 meV while the angular resolution is ≤0.5 degrees (∼0.021 Å−1 in k-space). All the measurements were carried out within 12 hours after cleavage and the temperature was kept constant at 50 K.