Origin of lattice compression of FeSe_1-xTe_x thin films on CaF₂ substrates

More than seven years have passed since the discovery of iron-based superconductors.^1,2 For the practical use, it is important to establish thin-film fabrication techniques, which we can utilize for producing not only low-power electric devices such as Josephson devices but also high-power electric cables constructed from coated conductors. In such superconducting thin films, highly oriented domain structures along both perpendicular and parallel directions to substrates are required to reduce weak link points at which superconductivity is depressed by external field. An epitaxial growth thus has been studied extensively so far. At an early stage of the study of thin-film growth, various perovskite-type oxide substrate materials were examined because of the similarity in lattice parameters.^3–10 However, most of the films grown on oxide substrates show lower T_C than the bulk crystals mainly due to severe diffusion of oxygen from the substrates, and only a very-thick film can exhibits a T_C comparable to bulk crystals. Based on these results, we proposed to use cubic CaF₂ (a₀ = 0.5463 nm) as a substrate instead of oxide substrates, and found that FeSe_1-xTe_x films exhibit higher T_C than bulk crystals and associated lattice compression along the in-plane direction, which shows a clear contrast to lattice expansion on many oxide substrates.^11–13 Thus, CaF₂ has been currently considered as one of the best substrate materials for iron-based superconductors. However, it has still remained as an open question why CaF₂ can induce lattice compression to FeSe_1-xTe_x despite its longer a-axis than that of FeSe_1-xTe_x. Moreover, a lattice-disordered layer with several nm thick is frequently observed at the interface, which should prevent the film on CaF₂ from being grown epitaxially. We have recently performed detailed comparison of the interface structure among three representative iron-based superconductors, SmAsFe(O,F) (1111), BaFe_2-xCo_xAs₂ (122), and FeSe_1-xTe_x (11),¹⁴ and found that the a-axis length of the grown film commonly shrinks regardless of the sign of lattice misfits between the films and CaF₂, which indicates the lattice matching is not the main reason of the lattice compression. In this paper, we have studied the interface between CaF₂ and FeSe_1-xTe_x thin films with respect to transmission electron microscope (TEM) and energy dispersive X-ray spectroscopy (EDX). We pay particular attention to the diffusion of constituent elements across the interface, especially the Te content (x) dependence.

All the films were prepared by pulsed laser deposition using a sintered target with stoichiometric composition as was described elsewhere.^9,11,15,16 CaF₂ (001) single crystals were used as a substrate, on which the films were grown at the substrate temperature of 300°C with their c axis normal to the substrate surface. We prepared five tetragonal FeSe_1-xTe_x films with different compositions (x = 0.0, 0.5, 0.6, 0.7, and 1.0.). Detailed specifications are summarized in Table I. These films were fabricated to a small thin plate for TEM observation using focused ion-beam as were previously reported.^14,17 TEM observation and EDX analysis were carried out using JEOL JEM2100F Thermal Field-Emission (FE) microscope.

Figure 1 shows cross-sectional TEM images near the interface of FeSe_1-xTe_x film and CaF₂ substrate, and corresponding electron diffraction patterns. The dark-field TEM images revealed that a thin layer of approximately 5 nm in thickness appears at the CaF₂ side from the interface similar to the previous results.¹⁷ While a part of this layer looks to keep crystalline structure, most of the area is disordered. We have already confirmed that the disordered layer shows up after FeSe_1-xTe_x is deposited on CaF₂,¹⁷ which suggests an occurrence of chemical reaction between them. Due to the presence of this disordered layer, it is hardly possible to trace how the epitaxial growth of FeSe_1-xTe_x begins from the (100) surface of CaF₂. Instead, we tried to see how the lattice parameters change across the interface from the corresponding electron diffraction patterns shown in Figs. 1(f)-1(j). Note that the direction of incident electron is not identical; Figs. 1(f), 1(h), and 1(j) (x=0.0, 0.6, and 1.0, respectively) are taken with CaF₂ [110] azimuth, while Figs. 1(g) and 1(i) (x=0.5 and 0.7, respectively) are taken with CaF₂ [100] azimuth. The diffraction patterns including CaF₂ [202] with FeSe_1-xTe_x [112] or CaF₂ [111] with FeSe_1-xTe_x [101] are shown in Fig. 2(a) for comparison among different x. One can see immediately that none of the films exhibits perfect lattice matching to CaF₂ but their a-axis lengths are always shorter than CaF₂. Note that for x = 0.0 lattice ordering of the films is much worse than the others e.g., diffraction spots from several c-axis domains are overlapped. For the other compositions, we obtained a single set of clear diffraction spots, which supports the c-axis orientation of the observed area of FeSe_1-xTe_x. Then, we can estimate the a-axis length of FeSe_1-xTe_x with referring to the diffraction spot from CaF₂, even though the direction of the incident electron is different. The results are summarized in Fig. 2(b) with the data for bulk crystals.^18,19 Since, the electron diffraction pattern for the x = 0.0 shown in Fig. 2(a) does not allow to estimate the a-axis length within a sufficient accuracy, we refer to the value for the film that were grown in the same way and has a similar film thickness.¹⁵ It is surprising from Fig. 2(b) that the a-axis length shows a good agreement between that for the x = 1.0 film and for bulk crystal, which strongly suggests that the FeTe film receives almost no strain effect from CaF₂. On the other hand, the decrease of a-axis length of the films becomes significant as x decreases, especially below 0.6. Even though the a-axis length of the bulk crystals decreases with decreasing x, this plot demonstrates that there is an “additional” decrease, and the deviation is enhanced with decreasing x. In other words, the additional shrinkage becomes remarkable with increasing the amount of Se in the film, while Te is inactive for this phenomenon. We note again that, even at x = 1.0, the a-axis length does not reach that of CaF₂, which supports that lattice matching is not a dominant factor to determine the a-axis length of FeSe_1-xTe_x.

In order to investigate the difference between the role of Se and Te in more detail, we performed EDX analysis across the interface, and compared among the films with different x. Figure 3 shows line profiles of the constituent elements with corresponding bright-field STEM images. The lower side is CaF₂ and the upper side is FeSe_1-xTe_x. In the four films of x = 0.0 – 0.7, we can identify characteristic white-band areas appearing at the interface, which is located inside the CaF₂ substrate. On the other hand, such a white-band area is not clearly appeared for FeTe, which suggests that the presence of Se gives a certain influence on the crystallographic ordering of the CaF₂ substrate. The line profiles actually imply that Se diffuses into CaF₂. In Fig. 3(a), Fe signal drops steeply at the interface between FeSe and the white-band area when tracing the signal of Fe from top to bottom, while the signal from Se remains even in the white-band area and starts decreasing after the white-band finishes. These results indicate that the white-band area is basically CaF₂ but contains Se in its matrix. Another finding is that the profile of Se and Te exhibits a finite shift as is commonly found in Figs. 3(b)-3(d). The signal of Te drops when it goes into the white-band area, while that of Se remains finite. This means that Te is almost absent in the white-band area. Such characteristic dispersion of Te can be confirmed even when Se is not present in the film. In Fig. 3(e), one can find for FeTe that the signal of Te drops and touches almost zero before that of Fe becomes zero near the interface, which again indicates that Te is hardly distributed into the CaF₂ substrate across the interface. Note that the white-band area can be found also in the dark-field image of FeTe/CaF2 film as shown in Fig. 1(e). We may, however, attribute it most likely to the F deficiency from the topmost layers of CaF2 substrate, as is indicated from Fig. 3(e).

It is interesting to see also that the signal of F changes complementary to that of Se. When going from the substrate to the film in Figs 3(a) – 3(c), the intensity of F signal starts decreasing even in the white-band area where that of Ca keeps sufficient intensity, which is typically observed for x = 0. This suggests that the F is deficient near the interface, and this F vacancy may induce the Se diffusion from the film. One may suspect that the F deficiency occurs at the surface of CaF₂ before FeSe_1-xTe_x is deposited, but this is not the case. We have confirmed using EDX that the surface of CaF₂ is stable and F deficiency does not occur after heating it at 300°C, a typical growth temperature, in a vacuum chamber. The F vacancy observed in CaF₂ shows up only after FeSe_1-xTe_x is deposited on it, which strongly suggests that a chemical property of Se and F plays an important role. In general, CaF₂ is considered to be a chemically stable material. However, our results suggest that it is not always the case.

The present results demonstrate on one hand that the Se diffusion from FeSe_1-xTe_x to CaF₂ occurs for x < 1.0, and on the other hand that the decrease of the a-axis becomes remarkable when decreasing x. Then it is natural to consider that these phenomena are strongly correlated. Based on our findings, we propose one possible scenario to explain the in-plane compression of FeSe_1-xTe_x thin films: 1) the initial several layers of FeSe_1-xTe_x react chemically with CaF₂ located underneath; 2) some portion of Se diffuse into CaF₂ substrate resulting the Se deficiency in the initial several layers; 3) the lattice of these layers shrinks by itself due to Se deficiency particularly along the a-axis direction; 4) the subsequent layers of FeSe_1-xTe_x are grown epitaxially on the initial layers with keeping the “shorter” a-axis length. We expect that the white-band area (disordered layers) works to isolate the effect of “epitaxy” from CaF₂, which suggests that the compressive strain is not directly due to the deformation of CaF₂ substrate. It should be noted that, in general, the effect of in-plane compression does not propagate perfectly all through the film thickness, and lattice relaxation occurs in thicker films. For the case of FeSe_1-xTe_x, the lattice relaxation occurs typically in the films thicker than 200 nm. We have already reported that the a-axis length of FeSe decreases first with increasing thickness up to 120 - 150 nm, and then turns to increase with further increase of the thickness.¹⁵ We believe that the variation of this a-axis length of FeSe is tightly related to the Se-concentration distribution along the thickness direction, which cannot be clarified even by the EDX analysis presented in Fig. 3, and we need to perform more quantitative analysis. From a practical purpose, the proposed mechanism gives the way to tune intentionally the a-axis length of FeSe_1-xTe_x on CaF₂ substrate by inserting FeSe layers with an appropriate thickness between CaF₂ and FeSe_1-xTe_x. As we suggested previously,¹⁹ it is important to tune independently the a-axis length and x for optimizing T_C. The present results offer one of the ways to realize the independent tuning of a and x. In the next step, we will utilize it on one hand for further quest of higher T_C in FeSe_1-xTe_x, and on the other hand for revealing a mechanism of possible very-high T_C in monolayer FeSe on SrTiO₃.^20–22 

We performed a microscopic observation of crystallographic ordering and atomic diffusion across the interface of FeSe_1-xTe_x superconductor thin films and CaF₂ substrates. When the film contains Se, a certain amount of Se diffuses into CaF₂ substrates resulting in excess Se vacancy and associated a-axis shrinkage in initial several layers of FeSe_1-xTe_x. On the other hand such a-axis shrinkage does not occur prominently in FeTe, indicating that Se and Te exhibit different behavior when being attached to CaF₂. In the present results, the decrease of the a-axis length is observed significantly in the film with its Te concentration x < 0.6. The perfect lattice matching to CaF₂ is never observed for the entire range of x, which means that the lattice parameter of CaF₂ is not a dominant factor to determine the lattice parameters of FeSe_1-xTe_x grown on it.

We thank M. Hanawa and S. Komiya for fruitful discussions. This research was partly supported by Strategic International Collaboration Research Program (SICORP), Japan Science and Technology Agency.