Anomalous He-Gas High-Pressure Studies on Superconducting LaO1-xFxFeAs

AC susceptibility measurements have been carried out on superconducting LaO1-xFxFeAs for x=0.07 and 0.14 under He-gas pressures to about 0.8 GPa. Not only do the measured values of dTc/dP differ substantially from those obtained in previous studies using other pressure media, but the Tc(P) dependences observed depend on the detailed pressure/temperature history of the sample. A sizeable sensitivity of Tc(P) to shear stresses provides a possible explanation.


Introduction
The discovery of superconductivity at temperatures as high as 26 K in LaO 1−x F x FeAs [1], a layered compound devoid of CuO 2 planes, has rekindled interest in hightemperature superconductivity. As for the cuprates [2], high pressure experiments can potentialy play an important role in furthering our understanding of the new Febased pnictides [3,4]. High-pressure experiments to 30 GPa by Takahashi et al. [5] on this compound for x = 0.11 using a solid (NaCl) pressure medium reveal that the superconducting onset in the electrical resistivity reaches temperatures as high as 46 K. The initial slope dT c /dP appears to increase with the fluorine content, as least for concentrations x ≤ 0.14 [3]. Resistivity and magnetic susceptibility studies by several groups [3,5,6,7] using various fluid pressure media agree that dT c /dP is positive initially, but differ widely on its magnitude. As in previous studies on the high-T c cuprates [2] and the binary compound MgB 2 [8], it would be of interest to carry out benchmark determinations of T c (P ) in magnetic susceptibility measurements on LaO 1−x F x FeAs using the most hydrostatic pressure medium known, He gas.
One notable result is that the non-superconducting, undoped pnictides LaOFeAs [3], CaFe 2 As 2 [9], SrFe 2 As 2 [10], and BaFe 2 As 2 [10] reportedly become superconducting under pressure when pressure media such as Fluorinert, methanol-ethanol, and silicone oil were used, superconductivity in CaFe 2 As 2 appearing at a relatively low pressure ≤ 0.4 GPa. However, recent dc susceptibility and electrical resistivity measurements by Yu et al. [11] on CaFe 2 As 2 using He-gas pressure medium fail to detect any sign of superconductivity to 0.7 GPa and 2 K. These authors suggest that not only in CaFe 2 As 2 , but also in SrFe 2 As 2 and BaFe 2 As 2 , non-hydrostatic stress components may have been responsible for the reported pressure-induced superconductivity. In addition, very recently Matsubayashi et al. [12] carried out parallel ac susceptibility and electrical resistivity measurements on BaFe 2 As 2 and SrFe 2 As 2 single crystals to 8 GPa pressure in a relatively hydrostatic cubic anvil pressure apparatus and find no trace of superconductivity in the former compound. In SrFe 2 As 2 bulk superconductivity, as evidenced by full shielding in the ac susceptibility, is only found in a rather narrow (2 GPa) pressure region centered about 6 GPa, whereas non-bulk (filamentary) superconductivity is revealed in the resistivity which falls to zero over a much wider pressure region. The potentially important role shear stresses play in the superconductivity of the Fe-based pnictides is emphasized in the recent review of Chu and Lorenz [4]. Indeed, shear-stress effects on T c (P ) are well known from studies on such diverse superconducting materials as organic metals [13], high-T c oxides [2], MgB 2 [8], and Re metal [14].
To our knowledge, the measurements of Yu et al. [11] on CaFe 2 As 2 are the only He-gas high-pressure studies of superconductivity yet carried out on an Fe-based pnictide. It would be of obvious interest to extend such studies to the LaO 1−x F x FeAs system to ascertain whether T c (P ) differs from the findings of earlier studies where less hydrostatic pressure media were employed. Using He as pressure medium brings a further benefit: it allows one to change the hydrostatic pressure at relatively low temperatures, thus permitting studies of phenomena impacting superconductivity which are both pressure and temperature dependent. An example for such phenomena in the cuprates are the well known oxygen ordering effects which have been shown in some systems to be the dominant factor determining the dependence of T c on pressure [2].
To throw some light on these issues, we have determined T c (P ) using He-gas pressure medium to 0.78 GPa for the original Fe-based superconducting pnictide, LaO 1−x F x FeAs, where x = 0.07 and 0.14. For these doping levels we find the initial pressure derivative dT c /dP to be positive, but markedly less than the published values using other pressure media; in addition, the dependence of T c on pressure is not reversible but depends on the detailed pressure/temperature history of the sample. Possible origins for this behavior are discussed.

Experiment
Polycrystalline LaO 1−x F x FeAs samples are prepared by solid-state reaction as described in previous publications [1,5]. For x = 0.14 both the previous resistivity measurements [3] and the present ac susceptibility studies are carried out on pieces taken from the same mother sample. The fluorine content in the present samples, x = 0.07 and 0.14, is determined from the lattice constant using Vegard's law; these samples have densities 6.704 g/cm 3 and 6.739 g/cm 3 [1] and masses 1.54 mg and 8.69 mg (or 5.65 mg), respectively.
To generate hydrostatic pressures as high as 0.8 GPa a He-gas compressor system (Harwood Engineering) was used in combination with a CuBe pressure cell (Unipress). AC susceptibility measurements at 0.1 Oe rms and 1023 Hz were carried out under pressure to the same high accuracy as measurements at ambient pressure by surrounding the sample with a primary/secondary compensated coil system connected to a Stanford Research SR830 digital lock-in amplifier via a SR554 transformer preamplifierthe. A two-stage closed-cycle refrigerator was used to cool the pressure cell to temperatures as low as 5-6 K; measurements were carried out upon warming up slowly through the transition at the rate ∼ 0.3 K/min. All susceptibility measurements were repeated at least once to verify that the reproducibility of the transition temperature was within 20 mK. Unless otherwise stated, the sample was cooled down to measure T c 30-60 min after a given change in pressure/temperature. Further details of the He-gas techniques are given elsewhere [15].

LaO 0.93 F 0.07 FeAs sample
For the x = 0.07 sample at ambient pressure the 80-20 transition width in the ac susceptibility is ∼ 1 K where T c ≃ 21.7 K from the transition onset and T c ≃ 21 K from the midpoint (see Fig. 1(a)). These values of T c are several degrees Kelvin less than those estimated from the resistivity onset at comparable fluorine concentrations [3].
Previous high-pressure resistivity measurements with Fluorinert pressure medium on LaFeAsO 1−x F x for x = 0.05, 0.08, and 0.11 yielded the following positive values of the initial derivative dT c /dP ≃ 2, 2, and 8 K/GPa, respectively, where T c was determined from the resistivity onset [3]. From these results one would anticipate that dT c /dP ≃ 2 K/GPa for x = 0.07. On the other hand, if T c is determined from the temperature at which the resistivity ρ drops to 0, one finds for x = 0.08 the initial pressure dependence dT c /dP ≃ 0.54 K/GPa, a value approximately 4× smaller than that from the resistivity onset (for x = 0.05 and 0.11 it was not possible to reliably estimate the temperature at which ρ → 0). In resistivity measurements, therefore, the value of dT c /dP obtained evidently depends on the criterion used to determine T c .
We now compare the values of dT c /dP obtained from the present ac susceptibility measurements using hydrostatic He-gas pressure on LaFeAsO 1−x F x to those obtained in the above resistivity studies. In Fig. 1(a) are shown our results for the F concentration x = 0.07. The large magnitude of the superconducting transition is consistent with bulk superconductivity; in fact, the shielding effect is approximately twice that expected for perfect diamagnetism. No correction is made here for diamgnetization effects. Under increasing He-gas pressure to 0.78 GPa, the superconducting transition is seen to shift monotonically to higher temperatures. Here T c is defined by the transition midpoint (see Fig. 1(a)); however, since the shape χ ′ (T ) of the transition does not change with pressure, the shift in T c with pressure is the same whether T c is defined from the transition midpoint or onset.
In Fig. 1(b) the dependence of T c on pressure is shown for all measurements on the x = 0.07 sample, the numbers giving the order of measurement. After the ambient pressure measurement (point 1), 0.78 GPa pressure is applied at room temperature (RT) to yield point 2. T c is seen to increase under pressure at the rate +1.20 K/GPa, clearly less than the value +2 K/GPa inferred from the resistivity onset but greater than the value using the ρ → 0 criterion at nearly the same F concentration (x = 0.08) [3]; in the resistivity studies the pressure was always changed at RT. The pressure was then successively reduced at low temperatures (62 K, 52 K, 45 K for points 2→3, 3→4, and 4→5, respectively) before cooling down further to measure T c . Up to and including point 4, the T c (P ) dependence is reversible; however, at ambient pressure (point 5) T c lies ∼ 220 mK lower than the initial value at ambient pressure (point 1).
Interestingly, after warming the sample back to RT and holding for 1.5 h, T c is seen to revert (point 6) back to its initial value. This behavior bears some resemblence to that observed previously from oxygen-ordering effects in the cuprates where T c (P ) may differ strongly whether the pressure is changed at RT or low temperatures [2]; there is, however, one notable difference --when pressure is reduced at low temperature in a cuprate with oxygen ordering, all values of T c would differ, and not just those below a certain pressure threshold.
To examine whether, as in the cuprates, there exists a particular (sub-RT) temperature above which such T c -relaxation occurs, we applied 0.66 GPa pressure at RT (point 7) and then released pressure at 60 K (point 8), reproducing exactly the previous results. Holding the sample at 100 K for 90 min resulted in no change in T c (point 9). However, after holding the sample at 200 K for 100 min, T c returned to its initial value (point 10). Warming back to RT for 1 h (point 11) and then for one week (point 12), resulted in no further change in T c . A pressure of 0.51 GPa was then applied at 60 K (point 13) and released at 50 K (point 14); these T c values faithfully track the +1.2 K/GPa straight line in Fig. 1(b).

LaO 0.86 F 0.14 FeAs sample
For the x = 0.14 sample at ambient pressure the 80-20 transition width in the ac susceptibility is ∼ 2 K where T c ≃ 13.7 K from the transition onset and T c ≃ 12.5 K from the midpoint (see Fig. 2(a)). This onset value is several degrees Kelvin less than from the resistivity onset (∼ 19 K); however, the temperature of the susceptibility midpoint is comparable to the resistivity zero point (∼ 12 K) [3].
Previous high-pressure resistivity measurements on LaO 0.86 F 0.14 FeAs, where pressure was always changed at RT, revealed that T c from the resistivity onset increases rapidly with pressure in Fluorinert pressure medium at the rate +12 K/GPa [3]. At a pressure of 0.66 GPa, therefore, T c would be expected to increase by approximately 8 K. In Fig. 2(a), however, the application of 0.66 GPa He-gas pressure at RT is seen to slightly broaden the transition in the ac susceptibility and shift it only slightly (∼ 0.2 K) to higher temperatures (point 2 in Fig. 2(b)), a shift 40× less than the 8 K expected! This difference in dT c /dP decreases to 15 × if the resistivity zero point is used (4.4 K/GPa).
The sample was then allowed to remain at RT for various cumulatitive lengths of time (20 h for point 3, 68 h for point 4, 112 h for point 5) for a total of 112 h, during which time the pressure at RT decreased only slightly to 0.65 GPa; surprisingly, T c is seen in Fig. 2(b) to decrease by ∼ 0.3 K (point 5)! Releasing the pressure then at 55 K to 0 GPa results in T c shifting further downwards to a temperature (point 6) ∼ 0.5 K less than its initial value at ambient pressure (point 1), the transition recovering its original sharpness. After the release of pressure at 55 K (point 6), holding the sample at 270 K for 1 h did not result in a further change in T c (point 7). The superconducting transition appears to be stuck at this lower value. Such a feature was not observed in oxygen ordering phenomena in the cuprates [2].
A second sample from the same synthesis batch was then studied to check these highly anomalous results, yielding the data points (open circles) labeled with primed numbers in Fig. 2(b). The value of T c at ambient pressure was identical to that of the previous sample. Applying 0.78 GPa He-gas pressure at RT shifted T c upward by only 0.23 K (point 2 ′ ), yielding a slope dT c /dP ≃ +0.30(1) K/GPa in excellent agreement with the results for the first sample (closed circles), but far less (40×) than that (dashed line) observed in resistivity studies by Takahashi et al. [3] using Fluorinert pressure medium. The pressure was then reduced successively at low temperatures to P = 0 (2 ′ → 3 ′ at 60 K, 3 ′ → 4 ′ at 55 K, 4 ′ → 5 ′ at 40 K, and 5 ′ → 6 ′ at 35 K). At ambient pressure T c now lies 0.62 K lower than the initial value (point 1 ′ ). Holding the sample for 1 h at 100 K caused no further change in T c (point 7 ′ ). T c was observed to shift upwards by 0.33 K after holding at 200 K for 1.3 h (point 8 ′ ), but no further shift in T c occurred after holding at 250 K for 1 h (point 9 ′ ) or at RT for 30 h (point 10 ′ ). A pressure of 0.33 GPa was then applied at 60 K (point 11 ′ ) and released again at 50 K (point 12 ′ ), yielding a value of T c approximately 0.8 K lower than the initial value (point 1 ′ ). The ambient pressure value of T c did not change further, even after holding at RT for 170 h (point 13 ′ )!

Discussion
In all previous high-pressure studies on the superconducting pnictides, pressure was changed at RT. We first compare the results of those studies to the present He-gas results for pressure change at RT. To our knowledge, the only measurements of T c (P ) under pressure on LaO 1−x F x FeAs for fluorine concentrations near those (x = 0.07 and 0.14) used in the present study are the resistivity measurements to 1.5 GPa with Fluorinert pressure medium by Takahashi et al. [3] for x = 0.05, 0.08, 0.11, and 0.14. As discussed above, if the pressure is changed at RT, the values obtained for dT c /dP from resistivity studies depend sensitively on the criterion used to determine T c ; the resistivity onset or ρ → 0 point give dT c /dP ≃ 2 or 0.54 K/GPa, respectively, in contrast to the present ac susceptibility studies using He-gas pressure where the intermediate value dT c /dP ≃ 1.2 K/GPa is observed. The difference in the value of dT c /dP is much larger at the higher fluorine concentration x = 0.14, where Takahashi et al. [3] find dT c /dP ≃ 12 K/GPa (onset) or 4.4 K/GPa (ρ → 0 point), the respective values being 40× or 15× higher than the 0.30 K/GPa found in the present Hegas experiments (see Fig. 2(b)). The fact that the anomalous temperature/pressure effects are most dramatic for the x = 0.14 sample, which lies near the substitution limit of F for O, suggests that the application of pressure may cause an irreversible phase separation. This would explain why the value of T c does not recover in our experiments after a pressure cycle, as seen in Fig. 2(b) We note that at the lower concentration x = 0.11 Takahashi et al. [3] report dT c /dP ≃ 8 K/GPa. On the other hand, at the same F-concentration the much lower value dT c /dP ≈ 1.2 K/GPa is obtained in a dc susceptibility measurement by Lu et al. [6] to 1 GPa pressure using an unspecified fluid pressure medium and in resistivity (onset) studies by Zocco et al. [7] using n-pentane : iso-amyl alcohol pressure medium to 0.94 GPa. It appears, therefore, that in the 1111 Fe-pnictides the pressure dependence of T c depends sensitively not only on the dopant concentration but also on which physical property is measured, how the value of T c is determined, and the type of pressure transmitting medium used. This would appear to support the view of Yu et al. [11] that shear stress effects play an important role in determining T c (P ) in the oxypnictides, large shear stresses generating significant changes in T c . The marked temperature/pressure history effects seen in Figs. 1(b) and 2(b) may be indicative of important shear-stress effects between grains in polycrystalline samples, even when purely hydrostatic He-gas pressure is applied. Parallel measurements on high quality single crystals would test this hypothesis. It is also possible that shortrange diffusion of oxygen or flourine anions within the crystal structure may occur in response to a change in pressure at RT, much as the "oxygen ordering effects" observed in the cuprate oxides [2].
In order to check whether the temperature/pressure history effects seen here might result from the penetration of the He pressure medium into the crystal lattice, we heated a 3.3 mg portion of the x = 0.14 sample used in the present He-gas experiments to 100 • C while connected to a sensitive mass spectrometer. We were unable to detect the slightest trace of He escaping from the sample. Subsequent vaporization of this sample in an ultra-sensitive mass spectrometer set the He impurity level at ∼ 1 ppm, an amount far too small to effect the dramatic changes observed in T c .
An alternative scenario is conceivable. The undoped compound LaOFeAs exhibits a spin-density-wave (SDW) and structural phase transition (tetragonal→orthorhombic) below 150 K [1,16]. Substituting O with F or applying pressure are believed to suppress this transition and allow a superconducting ground state to appear. A competition between a SDW instability and superconductivity is also observed in CeO 1−x F x FeAs [17]. Perhaps shear-stress effects result in superconducting and nonsuperconducting SDW regions in the sample which lead to the complex temperature/pressure history effects found in the present studies.
Whatever the explanation for the anomalous behavior of T c (P ), it is likely that in the oxypnictides, as in the cuprates, a full understanding of the manner in which T c changes under pressure may be difficult to obtain since it almost certainly depends on several factors simultaneously, including the strength of shear-stress effects as well as changes in the carrier concentration and the separation and area of the superconducting planes. Further experimentation is clearly needed here which focusses both on pressure-induced changes in the superconductivity and crystal structure on global and local scales.
pressure studies at Washington University were supported by the National Science Foundation through Grant No. DMR-0703896.