Introduction

More recently, the structurally simplest FeSe with space group (S.G.) P4/nmm has become very attractive since its Tc (8 K) can be drastically enhanced by a factor of 5 under external pressure1,2. Moreover, the single-layer FeSe film with huge superconducting gap (~20 meV) implies that the potential Tc could reach as high as 65 K even its bulk superconductivity (SC) decreases to ~40 K (ref. 3). Another high-Tc bulk SC derived from FeSe is intercalated AxFe2-ySe2 that is synthesized at high temperature (~1,300 K) but is only available for the large-sized alkali metals (A=K, Rb, Cs and Tl)4,5,6,7. The bulk Tc of AxFe2-ySe2 is ~30 K and the average crystal structure is a body-centred tetragonal phase (S.G. I4/mmm)4. However, the origin of SC and the precise superconducting composition are still under debate owing to the intrinsic phase separation and inhomogeneity in these materials8,9,10.

The low-temperature solution method is another way to approach the intercalated phase that is broadly applied and suitable for intercalating alkaline and alkali earth metals even those with small ionic radius. Many superconductors such as AxC60 and AxMNX (A=Li-K, Ca-Ba, Yb and Eu; M=Ti, Zr and Hf; and X=Cl, Br and I) were obtained through this method11,12. The relatively mild reaction keeps the host structure intact; therefore, the pure charge transfer without destroying the conductive layer might favour the higher Tc. Actually, the FeSe intercalates obtained from low-temperature alkali metal and NH3 co-intercalation exhibited a higher Tc of 30–46 K compared with the samples obtained from the high-temperature method13. One surprising feature is that small-sized cations, such as Li and Na, combined with the [NH2]− anions or NH3 molecules can be intercalated into the FeSe layers13. However, a closer examination of the discontinuity of separation between Fe layers (d), that is, the optimal Li (d≈8.3 Å), Na (d≈8.7 Å) and K (d≈7.4 Å) intercalates (Table 1), implies that the small cations may have more diverse coordination environments and arrangements of ligand groups compared with larger ones14,15. Furthermore, the fairly unstable NH3-rich phase Li0.56(1)N1.72(2)D4.63(3)Fe2Se2 with bilayers of NH3 molecules has been prepared at 250 K and its crystal structure is determined from the neutron powder diffraction pattern collected at 5 K (ref. 16). However, this NH3-rich phase has not been reported to intercalate with Na cations so far. These intricate complexations of alkali metals between the [Fe2Se2] interlayers imply that the cation intercalation potentially induces versatile intercalate especially for small alkali metals and need to be carefully investigated.

Table 1 Intercalated FeSe superconducting phases.
Full size table

In this communication, we identify three Na/NH3-intercalated phases from host FeSe through tuning the concentration of the Na-NH3 solution and subsequent post evacuation. The Tcs of 37, 45 and 42 K correspond to NH3-free Na0.65(1)Fe1.93(1)Se2 (phase I), NH3-poor Na0.80(4)(NH3)0.60(1)Fe1.86(1)Se2 (phase II) and NH3-rich phase (phase III), respectively. The NH3-free phase with ThCr2Si2-type structure is the first reported small alkali metal-intercalated iron-selenide without accompanying NH3 molecules. Moreover, the substitution of S for Se demonstrates that the structural stability and SC of this ThCr2Si2-type phase are more sensitive to the modulation of d than the NH3-poor intercalate.

Results

Multiple phases in Na-intercalated FeSe superconductors

Figure 1a shows the powder X-ray diffraction (PXRD) patterns for Na-intercalated FeSe phases prepared by immersion in 0.03, 0.1 and 0.3 mol l−1 Na-NH3 solutions. The phase I was obtained by immersion of FeSe in 0.3 mol l−1 Na-NH3 solution and subsequent post-evacuation treatment at ~260 K. As shown in Fig. 1a, there are three distinct phases judged by the differently strongest peak positions of (00l) diffraction that locate at 12.87°, 10.06° and 7.98° for the three phases, respectively. Fitting these patterns with highly preferred orientation, the separations between Fe layers (d) are 6.8339(2) Å, 8.7082(2) Å and 11.0721(3) Å. The schematic structure of the three intercalates with different NH3 contents and d values are shown in Fig. 1b. It can be seen that their d values increase from phases I to III patterns, indicating that the amount of intercalated cations or molecular groups between the FeSe layers also increases.

Figure 1: Three Na/NH3-intercalated FeSe phases.
figure 1

(a) The powder X-ray diffraction (PXRD) patterns of three intercalates with highly preferred orientation and host FeSe. The values of l are specified as 1 (primitive lattice) or 2 (body-centred lattice). (b) The schematic view of FeSe and three Na-intercalated phases with different separations between nearest Fe layers (d). (c) The magnetization curves of three intercalates measured with the zero-field-cooling (ZFC) and field-cooling (FC) modes at H=10 Oe.

Full size image

We analysed the chemical compositions of phase I–III compounds using an electron-probe micro-analysis (EPMA) and ion chromatography (IC) methods. The EPMA analysis shows that the composition of typical grains for phase I is Na0.65(1)Fe1.93(1)Se2. The difference in d values between phases I and II is 1.8625 Å that is consistent with that of AExHfNCl and AEx(NH3)yHfNCl (AE=Ca and Sr)17. Since the detected content of nitrogen is very small (~0.1), Na0.65(1)Fe1.93(1)Se2 is called the NH3-free phase. Phase II is obtained from 0.1 mol l−1 Na-NH3 solutions that is consistent with the reported intercalates13 and is called the NH3-poor phase. The results of EPMA and IC analyses indicated that the composition of phase II was Na0.80(4)(NH3)0.60Fe1.86(1)Se2, assuming that all nitrogen species came from NH3 molecules (Supplementary Fig. 1). The quite small difference in the Fe content between phases I and II may come from the reasonable uncertainties such as the highly chemical instability. It is noted that the formal iron valences of Na0.65(1)Fe1.93(1)Se2, +1.73(1), is significantly lower than normal Fe valence (+2) in FeSe. Here we also try to analyse the nitrogen species using Raman and Fourier transform infrared spectroscopy (FT-IR) techniques, but we could not identify whether the nitrogen species was an ammonia molecule or amide group (not shown). Therefore, the Fe oxidation state of NH3-contaning intercalates cannot be determined. Note that the phase II easily decomposes even under inert conditions (O2 and H2O concentration,<0.1 p.p.m.) at room temperature, which is attributed to the deintercalation of NH3 molecules from the interlayers. Furthermore, the difference in d between phases II and III is 2.369 Å, suggesting that phase III contains more NH3 molecules. The larger d of phase III could be attributed to the hydrogen bonding and additional intermolecular repulsion of NH3. Thus, it is called as the NH3-rich phase. It should be noted that this NH3-rich phase rapidly decomposes even at 250 K. It suggests that the NH3-rich phase is quite hard to purify by conventional methods similarly as the Li-intercalated counterpart16. Possible reaction schemes for the three intercalated phases are discussed in detail (Supplementary Discussion).

Figure 1c shows the superconducting transitions of three samples in the lower temperature range. Distinct superconducting transitions are observed, where the Tcs of 37 and 45 K correspond to the NH3-free and NH3-poor intercalates, respectively. The high superconducting volume fractions, 80–100%, and the M-H loops of 2 K (Supplementary Fig. 3) suggest that both phases are bulk SC. The magnetization curve of the NH3-rich phase shows two transitions of 45 and 42 K that are consistent with the features of mixture phases. Thus, the latter Tc should be assigned to the new NH3-rich phase. This fact reveals that the Tc is not simply proportionate to the separation between neighbouring FeSe layers18; the amount of charge transfer, FeSe4 tetrahedron distortion and interlayer coupling should also seriously influence the SC.

Crystal structure

The experimental chemical composition of NH3-free Na0.65(1)Fe1.93(1)Se2 indicates that each layer contains 3.5% Fe vacancies. The number of Fe vacancies is strikingly smaller than that of high-temperature synthesized KxFe2-ySe2 phase. On the other hand, the abnormal valence of Fe implies that Na0.65(1)Fe1.93(1)Se2 may be a metastable phase. To exactly determine the crystal structure of the NH3-free and NH3-poor compounds, we measured the PXRD patterns with a capillary assembled on a Mo-Ka anode diffractometer. The PXRD pattern was collected at room temperature and the Rietveld refinement result is plotted in Fig. 2a. The crystal structure of Na0.65(1)Fe1.93(1)Se2 is refined as an analogous KxFe2−ySe2 structural model4. The refinements smoothly converge to Rp=3.82%, Rwp=4.32% and goodness of fit (GOF)=1.90, respectively (Supplementary Table 1). The obtained crystal structure of Na0.65(1)Fe1.93(1)Se2 has a body-centred tetragonal lattice (S.G., I4/mmm) with a=3.7870(4) Å, c=13.6678(4) Å and V=196.03(8) Å3, reasonably smaller than those of AxFe2-ySe2 (A=K, Rb and Cs). The Fe-Fe distance of 2.6778(3) Å is comparable with those of other AxFe2-ySe2 phases. The Fe–Se distance is 2.4692(3) Å, which is the largest value among the pure FeSe as well as the intercalated phases. Meanwhile, the Se–Fe–Se angle of 100.13(2)° ( × 2) and 114.33(4)° ( × 4) indicates its FeSe4 tetragonal bears rather large distortion compared with other intercalated phases.

Figure 2: The powder X-ray diffraction (PXRD) patterns and their Rietveld refinements.
figure 2

(a) Results of the NH3-free Na0.65(1)Fe1.93(1)Se2 phase. This pattern comes from three phases: Na0.65(1)Fe1.93(1)Se2 (top, 91.3%), FeSe (middle, 3.5%) and Fe7Se8 (bottom, 5.2%). The inset shows the obtained crystal structure of ThCr2Si2-type Na0.65(1)Fe1.93(1)Se2. (b) The PXRD pattern and Rietveld refinement results for the NH3-poor Na0.80(4)(NH3)0.6Fe1.86(1)Se2 phase. This pattern comes from three phases: Na0.80(4)(NH3)0.6Fe1.86(1)Se2 (top, 95.4%), FeSe (middle, 2.4%) and Fe7Se8 (bottom, 2.2%). The inset shows the crystal structure of Na0.80(4)N0.6Fe1.86(1)Se2. The lower blue line shows the difference between experimental and calculated values.

Full size image

Furthermore, since the value of d in present NH3-poor phase is comparable to that of Li0.6N1.0D2.8Fe2Se2 (ref. 14), we tried to use the reported crystal structure of Li0.6N1.0D2.8Fe2Se2 to refine the PXRD pattern. The Rietveld refinement of the PXRD pattern of Na0.80(4)(NH3)0.60Fe1.86(1)Se2 intercalates is shown in Fig. 2b. Although the position of H cannot be obtained owing to the smaller scattering factor of hydrogen, we could determine the unique variable of the Se position (4e site: (0, 0, z)) in this high-symmetry unit cell (Supplementary Table 2). The refined lattice parameters are a=3.7991(2) Å, c=17.4165(4) Å and V=251.38(6) Å3, reasonably larger than those of Li0.6N1.0D2.8Fe2Se2. The Fe–Se bond length is 2.4110(2) Å, which is comparable to that of FeSe counterparts, indicating that presence of NH3 molecules weakens the structural change of FeSe layer.

Structure instability in Na-intercalated Fe(Se,S)

Our experiments show that the intercalation of the smallest alkali metal, Li, does not yield the NH3-free ThCr2Si2-type phase. This result suggests that the large d of [Fe2Se2] layers could not hold the small Li without combining [NH2]− anions or NH3 molecules. As previously reported19,20,21,22, the smaller alkali metals intercalated into [FeCuSe2] layers are more likely to form the CaAl2Si2-type structure with hexagonal close-packed structure instead of ThCr2Si2 type. Besides, it was found that the NH3-free ThCr2Si2-type phase cannot be formed when the solid solution FeSe1-zTez act as the host compounds. This fact indicates that the enlarged d in the host also cannot hold the small Na cation and destabilizes the ThCr2Si2 phase. It is likely that the Na0.65(1)Fe1.93(1)Se2 phase resembles the metastable iron pnictide superconductor Na1-yFe2-xAs2, which is on the edge of structural stability23. It is also known that the isovalent substitution of S for Se site could effectively tune the scale of d and the SC of FeSe24 and KxSe2-ySe2 (ref. 25). Therefore, S-substituted Na0.65(1)Fe1.93(1)(Se1-zSz)2 and NH3-poor Na0.80(4)(NH3)0.60Fe1.86(1)(Se1-zSz)2 samples were synthesized to investigate their structural and SC properties responses to the reduced d. Since the tetragonal phase of FeSe1-zSz vanishes as z rises above 0.5, the present work only focus on the low S range (0≤z≤0.5). The Tc of FeSe1-zSz first increases to 12 K and then rapidly decreases to zero (Supplementary Fig. 2), which is consistent with the previous report25.

The PXRD patterns of all S-substituted intercalated phases were measured (Supplementary Fig. 3). It can be seen that the main phase is an intercalated phase, and the peak position of (002) diffraction shifts to a higher angle as the sulphur content increases, indicating the contraction of the unit cell. The refined lattice parameters of both S-substituted samples are plotted in Fig. 3a. The linear decreases of a- and c axes are clearly observed as z increases up to 0.5, which unambiguously demonstrates that the S atoms are substituted into the Se sites. It can be seen that the solubility limit for Na0.65(1)Fe1.93(1)(Se1-zSz)2 phase is z=0.3, which is significantly smaller than that of the NH3-poor intercalates. Even when the host is tetragonal phase as z =0.4 and 0.5, we still cannot obtain ThCr2Si2-type phase, which only survives as the separation of Fe layer is larger than 6.6364(3) Å in Na0.65(1)Fe1.93(1)(Se0.7S0.3)2.

Figure 3: Lattice constants and magnetic susceptibility.
figure 3

(a) The lattice constants of Na0.65(1)Fe1.93(1)Se2 and Na0.80(4)(NH3)0.6Fe1.86(1)Se2 versus S substitution. The error bar represents 1 s.d. of uncertainty and is shown in the centre of symbol. The solid lines are guide to the eye. (b) The magnetization curves in the lower temperature region for two series of S substitution intercalates. For clarity, only the zero-field-cooling curves are shown.

Full size image

SC evolution in Na-intercalated Fe(Se,S)

Figure 3b plots the magnetization curves of S-substituted samples at a magnetic field of 10 Oe. As z increases up to a maximum, the Tc and shielding volume fractions of both intercalated phases are gradually suppressed by S substitution. For example, the Tcs are 25 K for both end members Na0.65(1)Fe1.93(1)(Se0.7S0.3)2 and Na0.80(4)(NH3)0.6Fe1.86(1)(Se0.5S0.5)2. The evolution of the SC and phase boundary of two intercalates, and host FeSe as the function of S content, are summarized in Fig. 4. It can be seen that the decrease of Tc for the two intercalates is rather mild and similar to the trend of Tc in KxSe2-y(Se1-zSz)2 phase, where Tc vanishes as the content of S rises to 0.8 (ref. 26). The percentage reductions of Tc for Na0.80(4)(NH3)0.60Fe1.86(1)(Se0.7S0.3)2 and Na0.65(1)Fe1.93(1)(Se0.7 S0.3)2 are 13% and 30%, respectively, indicating that the effect of S substitution for the NH3-poor phase is weaker than that for the NH3-free phase. In the present work, the formation of higher-sulphur intercalates is seriously hampered by the emergence of hexagonal phase in the host compounds. The fabrication of higher-sulphur tetragonal FeSe1-zSz with a soft chemical method is underway so as to fully understand the phase diagram.

Figure 4: The evolutions of SC and phase boundary.
figure 4

All of Tc were determined from the magnetization curves. Two different phase boundaries are clearly observed.

Full size image

Discussion

It is reported that the ThCr2Si2-type structure strongly competes with the CaAl2Si2 type, where the small alkali metal intercalation and the smaller d would destabilize the ThCr2Si2 type26. Moreover, if the Se was completely substituted by S through high-temperature treatment, the product would be the CaAl2Si2-type NaFe1.6S2. The phonon spectrum calculation of hypothetical ThCr2Si2-type NaFe1.6S2 indicates that negative vibration frequencies of Na atoms result in structural instability23. In present case, it is found that the Na-Na distance (~3.79 Å) in Na0.65(1)Fe1.93(1)Se2 is rather short than the counterpart (~3.86 Å) of CaAl2Si2-type NaFe1.6S2. With increasing the content of S, both Na-Na distances of intercalates further decrease to ~3.76 Å for z=0.3 and 0.5, respectively. The shortened Na-Na distance would inevitably increase Na-Na Coulomb repulsion, which may be the origin of structural destabilization. As z=0.3, the shortest Na-Na distance in Na0.65(1)Fe1.93(1)Se2 induces the strongest Na-Na repulsion, which possibly leads to the collapse of ThCr2Si2 structure. Therefore, our results show that the formation of the ThCr2Si2 phase for NH3-free Na-intercalated FeSe requires a stricter chemical environment and the unique treatment procedure compared with the NH3-containing intercalates.

Another point should be noted is the different Tc and their evolutions between host and intercalates. It is fact that the Tc, to some extent, will increase as d value increase from 5.52 to 8.71 Å. But Tc decreases slightly again as d increases above 8.71 Å. Actually, the summarized data on FeSe-based superconductors show that there is indeed an optimal d value, ~8.6 Å, below which the interlayer spacing is in proportion to Tc27. As d beyond this optimal value, other factors such as FeSe4 distortion, Se height off Fe layers and transferred charge from Na to FeSe layer, would seriously influence the value and evolution of Tc.

On the other hand, the suppression of SC by chemical pressure is quite similar to that of the pressured LiFeAs, where the shrinkage of unit cells leads to the monotonous decrease of Tc and shielding volume fractions28. Therefore, it is highly expected that the full-range S/Te-substituted intercalates should be prepared so as to thoroughly study the evolution of SC. Moreover, this ThCr2Si2-type Na0.65(1)Fe1.93(1)Se2 contains very few Fe vacancies and significantly differs from the highly Fe-vacant AxFe2-ySe2 obtained at high temperatures. We expect that the discoveries of intercalates with a nearly intact FeSe layer could promote the further understanding of ThCr2Si2-phase SC.

Methods

Synthesis

The high-purity powder precursors were synthesized using a modified high-temperature solid-state method. The typical process of FeSe and its solid solution uses iron granules (Alfa, 99.98%), selenium grains (Kojundo, 99.99%) and sulphur grains (Kojundo, 99.99%) that are put into alumina crucibles and sealed in silica ampoules. The samples were heated to 1,300 K for 30 h, then annealed at 673 K for 50 h, and finally furnace-cooled to room temperature. The Na pieces and Fe(Se, S) powders with nominal 1:2 mole ratio were loaded and sealed into a Taiatsu Glass TVS-N2 high-pressure vessel with a magnetic stirrer. These manipulations were carried out in an argon-filled glove box with an O2 and H2O content below 1 p.p.m. The vessel was taken out from the glove box and connected to a vacuum/NH3 gas line equipped with a turbo molecular pump and mass-flow controller. First, the vessel was evacuated using molecular pump and then placed in a bath of ethanol cooled by liquid nitrogen (~223 K). The ammonia cylinder and regulator were then opened allowing ammonia to condense into the vessel. Typically, 4–6 g of NH3 was condensed to form three different Na/NH3 concentration solutions. Further, the reaction vessel was closed and stirred for 3 h at 223–243 K. After the intercalated procedure finished, the vessel was opened and the solutions were evaporated at an ambient pressure. For NH3-free sample, the vessel was further evacuated to ~10−2 Pa using a molecular pump, whereas for other two NH3-containing samples, the evacuation process was not needed.

Structural and magnetic characterization

The PXRD patterns of products were measured by Bruker diffractometer model D8 ADVANCE with Cu-Kα anode (λ=1.5408 Å). To reduce the preferred orientation of PXRD patterns, the samples also were loaded into thin-walled capillary tubes and then measured using a Bruker diffractometer with Mo-Kα radiations (λ=0.7107 Å). The Rietveld refinement of patterns was performed using code TOPAS4 (TOPAS 2005, Version 3, Bruker AXS, Karlsruhe, Germany). The DC magnetization was measured by a vibrating sample magnetometer (SVSM, Quantum Design) at the low magnetic field of 10 Oe.

Chemical composition characterization

The chemical compositions of the samples were determined by EPMA with a backscattered electron mode. The real composition was determined as the average value of 10 points on a typical grain with dimensions of 50 × 30 × 5 μm. The Na and nitrogen contents in the samples were determined using the IC technique29. Typically, ~10 mg of the sample was dissolved in 5 mol l−1 hydrofluoric acid aqueous solutions and was diluted by adding water. The resultant solution containing Na+ and NH4+ was analysed by IC with a Shimadzu CDD-10A conductivity detector.

Additional information

How to cite this article: Guo, J. et al. Superconductivity and phase instability of NH3-free Na-intercalated FeSe1-zSz. Nat. Commun. 5:4756 doi: 10.1038/ncomms5756 (2014).