Main

The continuing search for new superconductors has recently yielded a new family of oxypnictides composed of alternating LaO1-xF x and FeAs layers1,2,3,4 with Tc values of 25–28 K, which can be raised to 40–43 K by replacing La with Ce (ref. 5) or Sm (ref. 6) or to 52 K by replacing La with Nd and Pr (refs 7, 8). Several experiments and band structure calculations suggest unconventional superconductivity in the paramagnetic Fe layer. First, ab initio calculations indicate that superconductivity originates from the d orbitals of what would normally be expected to be pair-breaking magnetic Fe ions, suggesting that new non-phonon pairing mechanisms are responsible for the high-Tc superconducting state11,12. Second, F-doped LaFeAsO is a semimetal, which exhibits strong antiferromagnetic fluctuations and a possible spin density wave instability around 150 K in the parent undoped LaFeAsO (refs 5, 13–17). And last, superconductivity may emerge on several disconnected pieces of the Fermi surface11,12,18,19, thus exhibiting the multi-gap pairing that has recently attracted so much attention in MgB2 (ref. 20).

Given the importance of magnetic correlations in the doped oxypnictides, transport measurements at very high magnetic fields are vital to probe the mechanisms of superconductivity. Indeed, first measurements of the upper critical magnetic field Bc2(T) have yielded a slope = dBc2/dT ≈ 2 T K-1 near Tc, for both La- and Sm-based oxypnictides2,3,4,5,6. From the conventional one-band Werthamer-Helfand-Hohenberg (WHH) theory21, such slopes already imply rather high values of the upper critical magnetic field at zero temperature Bc2(0) = 0.69Tc  ≈ 36 T for LaFeAsO0.89F0.11, ∼59 T for SmFeAsO0.9F0.1, and ∼72 T for PrFeAsO0.89F0.11, all well above Bc2(0) ≈ 30 T of Nb3Sn. However, studies of the high-field superconductivity in MgB2 alloys have shown that the upward curvature of Bc2(T) resulting from multiband effects can significantly increase Bc2(0), as compared to the WHH one-band extrapolation (see ref. 22 and references therein). Our high-field d.c. transport measurements on LaFeAsO0.89F0.11 samples show that Bc2(T) indeed exhibits signs of two-gap behaviour similar to that in MgB2 with a value of Bc2(0) that exceeds the WHH extrapolation by ∼2 times and which also exceeds the BCS paramagnetic limit, Bp (in tesla) = 1.84Tc (in kelvin) = 51.5 T for Tc = 28 K.

Polycrystalline LaFeAsO0.89F0.11 samples were made by solid state synthesis4. A sample ∼3 × 1 × 0.5 mm3 was used for our four-probe transport measurements in the 45 T hybrid magnet at the National High Magnetic Field Laboratory, supplemented by low-field measurements in a 9 T superconducting magnet. Our low-field data agree well with earlier data taken at ORNL on the same sample4, indicating its good temporal and atmospheric stability. The 45 T hybrid magnet was swept only from 11.5 T to 45 T owing to the static 11.5 T background field generated by the outer superconducting coil of the magnet combination, while lower fields were swept from 0 T to 9 T in a Physical Property Measurement System (PPMS) superconducting magnet with parameters shown in Fig. 1 legend.

Figure 1: Variation of LaFeAsO 0.89 F 0.11 sample resistance with applied magnetic field at fixed temperatures in the range 4.2 K to 25 K.
figure 1

a, The resistance R(B) for different temperatures taken in swept fields in the 45 T hybrid magnet at a measuring current of 1 mA and from 0 to 9 T in a superconducting magnet at a measuring current of 5 mA for B applied perpendicular to the broad face of the sample. The current density was ∼0.2 A cm-2 in the hybrid magnet measurements and 5 times higher in the PPMS superconducting magnet measurements. b, R(B) data taken in the hybrid magnet for B parallel to the broad sample face, which we believe has a higher fraction of a–b-oriented grains. The normal state resistivity at 30 K is estimated to be 0.15 mΩ cm with an uncertainty of ∼15%. The horizontal dashed lines indicate 10%, 50% and 90% of the resistive transition relative to the normal state resistance at the transition temperature, Rn(Tc), respectively.

Full size image

The results of our high-field measurements of the sample resistance as a function of magnetic field strength, R(B), are shown in Fig. 1. The broad R(B) transitions are not surprising because the sample consists of misoriented anisotropic crystalline grains. Given the predicted resistivity ratio Γ = ρc /ρab ≈ 10–15 for this layered compound11, the local Bc2(θ) ≈ Bc2(0)[cos2θ + Γ-1sin2θ]-1/2 should vary strongly, depending on the angle θ between the c-axis in the grain and the applied field. Thus, we can identify two characteristic fields: the high-field onset Bmax of the superconducting transition, and the zero-resistance, low-field onset Bmin, as illustrated by Fig. 1. The in-field transitions are shown in Fig. 1a for B perpendicular to the broad 1-mm-wide face of the sample, and in Fig. 1b for B applied parallel to the same face (in both cases, transport current was always perpendicular to B). Differences in the R(B) curves of Fig. 1 do indicate some grain texture, as the zero-resistance field Bmin in Fig. 1b (36 T at 4.2 K) is higher than in Fig. 1a (25 T at 4.2 K), suggesting that the plate-like grains tend to align their a–b planes parallel to the broad face of the sample.

Shown in Fig. 2a are the temperature dependences of the fields Bmin, Bmid and Bmax (all in T) evaluated at 10%, 50% and 90% of the normal state resistance at the transition temperature, Rn(Tc), respectively. The interpretation of Bmin and Bmax in polycrystals can be complicated by strong vortex pinning and the particulars of the grain misorientation distribution. However in our case, the analysis is simplified because: (1) the measured magnetization curves are nearly reversible, suggesting weak pinning; and (2) the crystalline anisotropy of the compound results in a broad distribution of the local Bc2(θ) values in different grains. Thus, Bmax(T) is associated with the larger in-plane upper critical field (T) ≈ (T)Γ1/2 because grains with their a–b planes oriented along the applied field become superconducting first upon cooling. In turn, the zero-resistance field Bmin(T) can be interpreted as the field below which those superconducting grains with Bc2(θ) > B form a percolative path at our low measuring current density ∼0.2 A cm-2. For Γ ≫ 1 and negligible thermal activation of vortices, both the effective medium and the percolation23 theories give Bmin(T) ≈ (T)/pc, where pc < 1 is a temperature-independent number determined by the percolation threshold defined by the material anisotropy and the orientational grain distributions of the specific sample. Accordingly, our interpretation of Fig. 2 is that Bmax(T) and Bmin(T) reflect the temperature dependences of (T) and (T), respectively, in the case when thermal activation of vortices is weak, as we now discuss.

Figure 2: Upper critical field–temperature phase diagram of LaFeAsO0.89F0.11.
figure 2

a, The measured fields Bmax(T) and Bmin(T) (black and red squares, respectively) along with the midpoint transition fields (blue circles). The filled and open symbols correspond respectively to the parallel and perpendicular field orientations. The dotted line shows the WHH curve defined by the slope of Bmin(T) at Tc. b, Bmax(T) (black squares) and Bmin(T) (red squares) plotted as functions of the reduced temperature T/Tc. The data points above 45 T were extracted by linear extrapolation of R(B) at B < 45 T to R(B) = 0.9Rn(Tc), as shown by dashed lines in Fig. 1. The lines correspond to Bc2(T) calculated from the two-gap theory for the parameters described in the text.

Full size image

The upward curvature of Bmin(T) in Fig. 2 might also be interpreted as the signature of the irreversibility field Bm(T) resulting from melting of a weakly pinned vortex lattice. Thermal vortex fluctuation effects are quantified by the Ginzburg parameter Gi = (8π2kBTcΛa2/ξcφ02)2/2 (ref. 24), where Λa is the London penetration depth, ξa is the a–b plane coherence length, φ0 is the magnetic flux quantum, ξc = ξaΓ-1/2 is the c-axis coherence length, and kB is the Boltzmann constant. The copper oxide high-temperature superconductors have strong thermal fluctuations resulting in Gi ≈ 1–10-2 and Bm(T) ≪ Bc2(T). The conventional low-Tc superconductors, in which vortex fluctuations are negligible, have Gi ≈ 10-6–10-10 and Bc2 - Bm ≪ Bc2. The coherence length ξc = [φ0/2πBc2(0)Γ1/2]1/2 can be estimated for Γ = 15 and (0) = 60 T in Fig. 2 as 1.2 nm, which for Tc = 26 K and Λa = 215 nm, extracted from recent NMR measurements25, yields Gi = 3.4 × 10-4, a value close to Gi = 2.1 × 10-4 of clean MgB2, but a value some 30 times smaller than Gi for the least anisotropic high-temperature superconductor, YBa2Cu3O7-x, leading us to classify LaFeAsO0.89F0.11 as an ‘intermediate-Tc’ superconductor. We estimate Bm based on a theory of vortex fluctuations in moderately anisotropic superconductors24, which shows that Bc2(T) and Bm(T) at T < 0.5Tc differ only by ∼20%. Therefore, we conclude that the temperature dependence of the resistive transition field Bmin(T) at T < 0.5Tc reflects the behaviour of (T) rather than that of the melting field Bm(T).

As is evident from Fig. 2, (T) exhibits a significant upward curvature, which is much less pronounced for (T). Because this behaviour is similar to that observed in dirty MgB2 films22, we suggest, in agreement with the ab initio calculations11,12, that superconductivity in LaFeAsO0.89F0.11 results from two bands: a nearly two-dimensional electron band with high in-plane diffusivity D1 and a more isotropic heavy hole band with smaller diffusivity D2. The upward curvature of Bc2(T) is then controlled by the ratio η = D2/D1: for η ≪ 1, the upward curvature is pronounced, while for η ≈ 1, Bc2(T) exhibits a more traditional WHH-like behaviour21. For fields along the a–b plane, the parameter η should be replaced with η = D2/[ ]1/2, allowing strong anisotropy to significantly increase η for B||ab as compared to B||c (ref. 22). In this case the upward curvature of (T) does become less pronounced than for (T), in agreement with the data in Fig. 2.

To check the self-consistency of our interpretation, we fit (T) in Fig. 2, using the two-gap theory. We took η = D2/D1 = 0.08 and the interband BCS coupling constants λ12 = λ21 = 0.5, but the results are relatively insensitive to the choice of either λ12 and λ21 (including their sign) or the intraband coupling constants λ11 and λ22. For example, Fig. 2 shows a rather good fit for the case of strong interband repulsion λ12 λ21 ≫ λ11 λ22 suggested in ref. 12. However, this theory also shows that the fit remains nearly as good, even if we assume that intraband pairing is significant, λ11λ22 > λ12λ21, with all coupling constants being of the same order of magnitude. Thus, our experimental data do not yet enable us to unambiguously distinguish between different pairing scenarios suggested in the literature, but they do indicate a significant difference in the effective masses in the electron and hole bands. For B||ab, we rescaled the parameter η → ηΓ1/2 ≈ 0.31, taking the estimate Γ = ρc/ρab ≈ 15 suggested in ref. 11, which describes (T) well, as is also evident from Fig. 2. Therefore the two-gap scenario is qualitatively consistent with our experimental data.

The newly discovered LaFeAsO0.89F0.11 exhibits exceptionally high Bc2, and this obviously non-optimized material has been quickly synthesized in bulk form showing zero resistance above 35 T at 2.5 K. Moreover, given the high dBc2/dT values of 2–3 T K-1, which have also been observed in the Sm-based (Tc = 43 K; ref. 6) and Pr- or Nd-based oxypnictides (Tc = 52 K), it seems reasonable to expect (T) and (T) values 1.5–2 times higher than LaFeAsO0.89F0.11, which puts (0) above 100 T. Thus doped oxypnictides appear as a new family of high-field superconductors, for which extensive pulsed-field measurements in the 100 T range will be required to fully reveal the novel physics of competing superconducting and magnetic orders. It is tempting to think that they may also have great importance for high-field applications.