Emerging superconductivity hidden beneath charge-transfer insulators

In many of today's most interesting materials, strong interactions prevail upon the magnetic moments, the electrons, and the crystal lattice, forming strong links between these different aspects of the system. Particularly, in two-dimensional cuprates, where copper is either five- or six-fold coordinated, superconductivity is commonly induced by chemical doping which is deemed to be mandatory by destruction of long-range antiferromagnetic order of 3d9 Cu2+ moments. Here we show that superconductivity can be induced in Pr2CuO4, where copper is four-fold coordinated. We induced this novel quantum state of Pr2CuO4 by realizing pristine square-planar coordinated copper in the copper-oxygen planes, thus, resulting in critical superconducting temperatures even higher than by chemical doping. Our results demonstrate new degrees of freedom, i.e., coordination of copper, for the manipulation of magnetic and superconducting order parameters in quantum materials.

itself is advantageous in achieving homogenous oxygen configurations. Thin films of Pr 2 CuO 4 have been grown intentionally at temperatures lower than optimal as a reduced crystallite dimensions are advantageous for a homogeneous annealing experience.
In general the annealing process in oxide materials is a diffusion process. In particular, regular O(1) and O (2), as well as apical O (3) sites are occupied or evacuated in the T9-cuprates. Here, we use a high precision partial oxygen pressure monitoring and control system (POPMCS) combined with X-ray diffraction and transport data of Pr 2 CuO 4 for the analysis of the annealing process. A commercial quartz tube furnace equipped with a turbo molecular pump and POPMCS was used. The Pr 2 CuO 4 film was mounted on the tip of a SSA-S alumina tube placed at the center of the quartz tube.
Starting from the standard annealing process (Fig. 1) typically applied to electron doped cuprates 25,26 , we split the annealing procedure, thus a two-step annealing process. Our systematic investigation on this new two-step annealing scheme reveals that only certain annealing conditions are suitable to preserve the T9-structure and induce superconductivity into Pr 2 CuO 4 . In Fig. 1b, we plot the electronic phase diagram of Pr 2-x Ce x CuO 4 (standard annealed), where the superconducting transition temperatures of 273 c-axis oriented, single phase thin films of Pr 2-x Ce x CuO 4 on (001) SrTiO 3 (a 5 3.905 Å ) substrates are shown for 0.00 , x , 0.25. In contrast to the phase diagram for standard annealed Pr 2-x Ce x CuO 4 , the ex-situ two-step annealing process allows superconductivity even without cerium. The phase diagram shown in Fig. 1c shows that superconductivity appears at all doping levels up to x < 0.22 and the highest T c is not at x 5 0.15 but 0.00, in stark contrast to the commonly observed phase diagram (Fig. 1b). The newly obtained superconducting phase diagram indicates that the apparent symmetry of electronic phases for hole and electron doped cuprate superconductors with respect to the antiferromagnetic-insulating ground state might be an artifact of commonly used annealing treatments, thus, not representative. Instead, it appears that for zero doping, only the T-phase is an antiferromagnetic Mott insulator whereas the T9-phase is a superconductor, in agreement with the first principle methods' predictions [2][3][4] .

Discussion
Comparing the influence of doping to the influence of annealing to Pr 2 CuO 4 reveals that a hidden, hole-like Fermi surface may be present. The Ce doping dependence of the evolution of the Fermi surface of Nd 2-x Ce x CuO 4 has been reported by Armitage et al. 27 for x 5 0.04, x 5 0.10 and x 5 0.15. Traces (small but finite density of states) of a hole-like Fermi surface can be detected even for x 5 0.04 27 . However, such a sample is neither metallic nor superconducting owing to the annealing conditions applied. The hidden Fermi surface suggests that the applied annealing conditions were not optimal. Commonly, the observed Hall coefficient 14 is negative for x 5 0.04. The negative Hall coefficient can be attributed to ''hot spots'' located at (p, 0) and (0, p). The overall contribution to the Hall coefficient of those hot-spots is significant for Pr 2-x Ce x CuO 4 as the Hall coefficient is negative up to x < 0.17 (Fig. 2b). The Hall coefficients R H taken on superconducting Pr 2 CuO 4 show unambiguously that the origin of metallic conduction and superconductivity itself is not electron doping but points towards a redistribution of spectral weight from those anti-ferromagnetic ''hot-spots'' into the hole-like Fermi-surface 25 . Figure 2(c) shows the temperature dependence of R H for superconducting Pr 2 CuO 4 , and the Hall coefficient of standard annealed logarithmic and reciprocal scaling is used for the oxygen pressure and the absolute temperature, respectively. Thermodynamic stability lines for the copper-oxygen system and Pr 2 CuO 4 52 are shown. Pr 2 CuO 4 films were grown using a radio-frequency activated oxygen plasma (O*) by molecular beam epitaxy. The oxygen pressure during the synthesis is 2 3 10 26 Torr, corresponding to an equilibrium molecular oxygen pressure of 10 6 Torr. The synthesis temperature of Pr 2 CuO 4 is 650-750uC. Standard annealing is carried out at temperatures between 550 and 650uC under 10 29 Torr. In the two-step annealing process 53 , Pr 22x Ce x CuO 4 is annealed ex situ first at 750-850uC and 7.6 3 10 22 Torr O 2 and subsequently annealed at temperatures between 450 and 700uC under high vacuum. In (b), the doping dependence of the superconducting phase diagram of Pr 2-x Ce x CuO 4 is shown for 273 different samples obtained by the standard annealing process. For 0.00 , x , 0.10, Pr 2-x Ce x CuO 4 is an antiferromagnetic insulator (AFI). For 0.11 , x , 0.23, superconductivity is induced by the standard annealing process with a maximum T c of 25 K at x 5 0.14. In (c), the doping dependence of the superconducting phase diagram of Pr 2-x Ce x CuO 4 is shown. Data points (black triangle) have been taken from 16 . At x 5 0.00, results of 84 samples are summarized (star). Dashed lines represent the phase diagram as obtained in Fig. 1(b). : the additional oxygen acts as a dopant for holes and in a similar way as Sr doping. We like to highlight the fact that the temperature dependence of the Hall coefficient is not that of a simple metal but rather demonstrates the competition between a hole-like metal and an antiferromagnetic insulator. This asymmetric scenario 1,33-35 between square-and octahedral coordinated cuprates also shows that their electronic correlations are entirely different 3 . The absence of a doping mechanism in our elaborate annealing process is independently supported by the fact that the in-plane lattice constants of as-grown and annealed Pr 2 CuO 4 films are constant upon annealing as it is well known that electron-doping stretches and holedoping shrinks the Cu-O bonds in the CuO 2 planes due to accumulation or depletion of electrons to/from the Cu-O dps anti-bonding bands 36 . The presence of additional oxygen in as-grown Pr 2 CuO 4 is well established 23,24 as is its removal by annealing. We visualized our annealing scenario in Fig. 3a. The as-grown crystal contains more than the stoichiometric amount of oxygen which are randomly distributed at apical sites (Fig. 3a). After the first annealing step we find that the lattice parameters are nearly unchanged (Fig. 3f) when compared to the as-grown sample (Fig. 3e). However, its resistivity value is significantly higher (Fig. 3c). We explain such behavior by the introduction of oxygen vacancies in the CuO 2 plane since such defects would disturb electronic conduction severely. The second annealing step does repair those in-plane defects by relocating apical oxygen atoms to the planes and consequently the resistivity is lowered significantly (Fig. 3d). This final step creates a situation similar to what has been observed after an annealing treatment 37 for the cerium doped superconductors 38 . Overall we do observe that the caxis lengths decreases upon annealing (Fig. 3 e-g) and that has been unambiguously proven to be associated to the removal of apical oxygen by neutron scattering 23,24 . A typical value of the oxygen offstoichiometry estimated from neutron scattering experiments of asgrown Nd 2 CuO 4 1 d single crystals is d < 0.05, which indicates that one Cu ion out of ten unit cells is pyramidal coordinated. Experimentally, this is a sufficient condition to stabilize a long-range antiferromagnetic order even at Ce doping levels of x 5 0.15 39 . In Ref. 39 it was shown that even for x 5 0.15 the as-grown cuprate is an antiferromagnetic insulator with a T N 5 150 K. After annealing, however, the cuprate system goes into the superconducting state. The only chemical difference is that occupied apical oxygen sites have been evacuated during that annealing process. Those occupied apical oxygen sites break the symmetry for all nearest and nextnearest-neighbor Cu plaquettes. Such a locally broken symmetry localizes electrons primarily on one Cu site and induce a gap in the Fermi surface. Therefore, the doping process in electron doped cuprates might be considered as a band filling process, as its ground state is already a metal 40 .
It is worth mentioning that the entire annealing process is a diffusion process as long as thermodynamic limits are not violated. Certainly, those limits have been violated considering earlier reports 41 . In contrast to the standard annealing process applied for bulk specimens, thermodynamic constraints, e.g., the Pr 2 CuO 4 Pr 2 O 3 1 Cu 2 O stability line, may not be crossed in our 2-step annealing process. As for the standard annealing process, reduction conditions below the thermodynamic stability regions may harm the T9 phase, therefore RE 2 O 3 oxides are often observed and consequently cause an increase of the absolute resistivity value. The annealing conditions applied in the first annealing step of our experiments are above the thermodynamic stability lines of Pr 2 CuO 4 and CuO, thus, decomposition products, i.e., Pr 2 O 3 , can be ruled out in contrast to other experiments as we do not see indication of their presence either by transmission electron microscopy or X-ray diffraction. Besides the influence of the annealing conditions on the electronic transport properties (Fig. 4a,b), the crystallite dimensions of the thin film are also affected. Low annealing temperatures result in larger (Dq x ) 21 values (Fig. 4c and 4d), though the superconducting transition temperatures are constant (Fig. 4a, 4b). Both of the annealing steps of our two-step annealing process are not independent and their correlation to superconductivity is visualized in Fig. 4e where the superconducting transition temperature (T c ) is plotted as a function of the first-(T a ) and second-(T red ) annealing temperatures. For optimal superconducting transition temperatures, a low T a requires a low T red and a high T a requires a high T red . Consequently, when the annealing time and the oxygen partial pressures are kept constant, optimal superconducting transition temperatures are associated to T a and T red in an arc shaped relation.
Finally, we compare our data to results reported from first principle calculations mentioned earlier [2][3][4] . The contrasting ground states in square-planar and octahedral coordinated cuprates, i.e., T9 and T, are consequences of the difference in the charge-transfer gap D 0 , originating primarily from the different oxygen coordination. Vacant apical sites substantially reduce the electrostatic potential at the copper site, thus, the 3d 9 Cu energy levels of the T9-phase are lower than in the T-phase, whereas the 2p 6 O energy levels remain almost constant 42,43 . A simple evaluation of the unscreened D 0 from Madelung potential calculations 44,45 show that the difference in D 0 between T9-and T-phases is in the range of several eV -therefore, the charge transfer gap might be very small or may even vanish in the T9cuprates. Under such circumstances, the model of ionic binding, which is tacitly assumed in the discussion of the charge-transfer energy, loses its vindicability. Instead, hybridization effects between Cu 3d x 2 -y 2 and O 2p xy orbitals may dominate electronic correlations, though they are not taken into account in the commonly used t-J model 46 . A superconducting ground state in square planar coordinated cuprates, where doping is not a prerequisite but an option, may promote a deep understanding of the rich variety of electronic phases of cuprates as they depend on coordination, doping and diluted impurities 47 . Moreover, the new phase diagram of square-planar coordinated cuprates implies the following question: Does T c further increase upon hole-doping? A recent article by Takamatsu et al. 48 indeed observed superconductivity in hole doped square-planar coordinated cuprates. Answering may provide a fundamental understanding of the mechanism of high temperature superconductivity. Certainly, the induction of a long range commensurate 3D antiferromagnetic order by a tiny amount of apical oxygen in T9-cuprates demand for a thorough analysis outside of the commonly successful theoretical treatments. As the competition of antiferromagnetic and superconducting order in T9-cuprates ultimately tunes the electronic properties, e.g., r(T), R H (T), a microscopic understanding would be beneficial. The possible solution for a quantitative analysis of site specific occupancies of oxygen in T9-cuprates is either via neutron scattering experiments (bulk samples) or 17 O nuclear magnetic resonance (NMR) spectroscopy 49 .

Methods
Thin films of c-axis oriented, single phase Pr 2 CuO 4 were epitaxially grown on (001) SrTiO 3 (a 5 3.905 Å ) substrates by molecular beam epitaxy (MBE). The growth of the T9-Pr 2 CuO 4 films was performed in a custom-designed MBE chamber 50,51 (base pressure , 10 29 Torr) from metal sources by using multiple e-gun evaporators and an atomic oxygen source (0.5 sccm, radio-frequency (RF) power of 250 W) as an oxidizing agent. The cation stoichiometry was adjusted by controlling the evaporation beam flux of each constituent element by electron impact emission spectrometry (EIES) (Guardian IV, Inficon, USA) via feedback loops to the e-guns. Ultra-fine tuning of the evaporation beam fluxes (6 0.005 Å /s) was done by reflection highenergy electron diffraction (RHEED) monitoring 51 . Typically, the substrate temperature for the growth of T9-Pr 2 CuO 4 thin films was T s 5 600-650uC. The film thickness is 1000 Å . For comparison purpose, some of the films were reduced in-situ after the growth under the ultra-high vacuum (UHV) environment. accordance to the results deduced from neutron and X-ray scattering analysis and electronic transport data. (a) In the as-grown state, random apical sites of copper are occupied (apical oxygen). During the first annealing step of our two-step annealing procedure, not apical but regular oxygen sites of the CuO 2 planes are being evacuated. During the second annealing step, the defective CuO 2 plane is being ''healed'' by an oxygen rearrangement from the apical sites to regular in-plane sites (shrinkage of c-axis). (b)-(d) Evolution of r(T) characteristics and lattice constants after each synthesis step. The asgrown T9-Pr 2 CuO 4 thin film is insulating and the optimally reduced films (after step II) are superconducting while r(300 K) is reduced by more than 2 orders of magnitude. The T9-Pr 2 CuO 4 thin films just after step I are even less conductive than the as-grown ones. Using the MBE-grown films, we investigated the reduction condition dependence of the properties of T9-Pr 2 CuO 4 . A commercial quartz tube furnace of 60 cm length and 30 mm diameter was used. The furnace is equipped with a turbo molecular pump (TMP) and a commercial (SiOC-200, STLAB, Japan) high precision partial oxygen pressure monitoring and control system (POPMCS). The POPMCS allows a precise control of the oxygen partial pressure between 10 21 to 10 216 atm by mixing an inert gas, e.g., N 2 , and oxygen at an electrochemically controlled oxygen diffusor (yttrium stabilized zirconium oxide). The Pr 2 CuO 4 film was mounted on the tip of a SSA-S alumina tube placed at the center of the quartz tube in longitudinal direction. Prior to its first usage the quartz tube was cleaned in boiling piranha clean whereas the alumina tube was rinsed by deionized water. The cleaned quartz tube and SSA-S alumina tube were prebaked at 1000uC for 10 h under ultra-high vacuum. Prior to the first annealing step, the partial pressure of oxygen was adjusted to a defined value. The N 2 /O 2 gas mixture was kept at a constant flow rate of 500 sccm throughout all experiments. The second annealing step is performed in the same tubular furnace evacuated in 10 25 Torr residual gas pressure. Figure 4 | The temperature dependence of the resistivity (a,b), their associated high resolution reciprocal space maps (HRRSM) of fully relaxed Pr 2 CuO 4 films grown on (001)SrTiO 3 substrates (c,d), and the relationship between the first (T a ) and second (T red ) annealing temperature and their influence on the superconducting transition temperature T c (e). In (a,c), a Pr 2 CuO 4 film was annealed at T a 5 750uC and 7.6 3 10 22 Torr oxygen for 1 h (first annealing step) , followed by a reduction process at T red 5 450uC under high vacuum for 10 min. The electronic transport shows metallic behavior with a superconducting transition at 26.0 K and a residual-resistivity-ratio (RRR) 5 7. The relative position of the (2109) diffraction spot of Pr 2 CuO 4 to the (2103) SrTiO 3 diffraction spot shows that Pr 2 CuO 4 films are epitaxial but relaxed grown on (001) SrTiO 3 . The in-plane lattice constant of the Pr 2 CuO 4 films is 3.96 Å . (Dq x ) 21 < 80 nm provides a rough estimation of the lateral crystallite dimensions. In (b, d), a Pr 2 CuO 4 film was annealed at T a 5 850uC and 7.6 3 10 22 Torr oxygen for 1 h (first annealing step), followed by a reduction process at T red 5 650uC under high vacuum for 10 min. The electronic transport shows metallic behavior with a superconducting transition at 25.0 K and RRR . 5. The relative position of the (2109) diffraction spot of Pr 2 CuO 4 to the (2103) SrTiO 3 diffraction spot shows that Pr 2 CuO 4 films are epitaxial but relaxed grown on (001) SrTiO 3 . The in-plane lattice constant of the Pr 2 CuO 4 films is 3.96 Å . (Dq x ) 21 < 250 nm provides a rough estimation of the lateral crystallite dimensions. The influence of the annealing history on the superconducting transition temperature T c is given in (e). Here, the oxygen partial pressures during the first and second annealing steps were kept constant and are 7.6 3 10 22 Torr and high vacuum, respectively. T c levels as high as 26.0 K can be reached for Pr 2 CuO 4 films grown on (001) SrTiO 3 substrates. www.nature.com/scientificreports