Is superconductivity in the iron arsenides conventional? The large isotope effect on both the magnetic and superconducting transitions may indicate that magnetic fluctuations are involved in the superconducting pairing.
Nearly half a century passed between the discovery of superconductivity by H. Kamerlingh Onnes and its theoretical description by Bardeen, Cooper and Schrieffer (BCS theory) in 1954. The isotope effect — the change in the superconducting transition temperature, Tc, caused by a change in the ion mass — was a strong indicator that phonons were responsible for superconductivity, and the discovery of a large isotope effect in mercury was one of the catalysts leading to the BCS theory. Bardeen, Cooper and Schrieffer showed that the 'glue' that binds together electrons to form Cooper pairs in the superconducting state is indeed lattice vibrations. So far, no mediating agent other than phonons has been definitively shown to occur in any superconductor; however, there is strong evidence that in the copper oxides and other unconventional superconductors magnetism might be a factor. Writing in Nature, Rong Hua Liu et al.1 report measurement of the iron isotope effect coefficient (IEC) in the new iron arsenide superconductors and argue that the superconductivity might be mediated by magnetic fluctuations.
Just over a year ago, a new variety of superconducting material containing iron arsenide layers was discovered2. In the FeAs layer, iron atoms sit on a square two-dimensional lattice with arsenic layers above and below arranged such that the iron atoms are all tetrahedrally coordinated. Depending on the nature of the layer separating the FeAs layers, transition temperatures of up to 55 K are possible3. As in the layered copper oxides, the FeAs layer must be 'doped' to become superconducting, usually by ion substitution in the separation layer. In both of these classes of superconductor, the undoped material is magnetic. On doping, the ordered magnetic state is destroyed and superconductivity sets in. Unlike in the copper oxide materials, it is possible in the same iron arsenide compound to dope either electrons or holes into the FeAs layer to induce superconductivity, as both electron and hole Fermi surfaces are present.
A great deal of information has been collected on these materials in a relatively short time, but, as for the copper oxides, there is still no clear-cut agreement over the mediating agent, although there are strong indications that magnetism may have an important role. It has been shown that the order parameter is not a simple s-wave state but is unconventional and changes sign between different parts of the Fermi surface.
Liu and co-workers1 measure the iron IEC for both the magnetic (TSDW) and superconducting transition temperatures in a hole-doped material (SmO1–xFxFeAs, which is of type 1111; Fig. 1) and an electron-doped material (Ba1–xKxFe2As2, which is of type 122). The IEC for the magnetic transition is measured in the undoped (x = 0) compounds and that for the superconducting transition is measured in the doped materials (x = 0.4 for potassium and x = 0.15 for iron). What they find is somewhat surprising: the IECs (α) for the magnetic and superconducting transitions in both compounds are essentially the same: 0.37(3) = α = −d(ln T)/d(ln M), where T is either TSDW (spin-density-wave temperature) or Tc and M is the iron mass. Moreover, the authors show that on substitution of a heavier iron isotope, the magnetic transition temperature, TSDW, is suppressed; modifying the lattice phonon modes affects the magnetism. This observation experimentally confirms that the lattice phonons and magnetism are intimately coupled in the iron arsenides.
On cooling the undoped iron arsenide materials, the nearest-neighbour iron moments order ferromagnetically or antiferromagnetically into a stripe pattern with alternating rows of iron atoms ordered antiparallel. The ferromagnetic ordering along the stripe direction and antiferromagnetic ordering perpendicular to the stripes lead to a frustrated magnetic system with the frustration being ultimately lifted by an orthorhombic distortion of the plane. Unlike the copper oxides, the undoped system is metallic and the magnetic ordering is an itinerant spin-density-wave (SDW) state. Experimentally, the magnetic SDW ordering and phase transition occur nearly simultaneously for the 122 systems and within several kelvin of each other for the 1111 system, indicating a strong coupling between the ordered magnetism and lattice. Calculations4 have also shown that a large magnetoelastic coupling effect exists and strongly indicates that the tetragonal–orthorhombic phase transition is most likely driven by the magnetic transition.
As the materials are doped at low temperatures from the magnetic orthorhombic phase, the material becomes non-magnetic, tetragonal and superconducting. The order of these transitions is still not completely clear for all of the various materials, as the synthesis of accurately doped compounds is difficult. Superconductivity is thought to occur only in the non-magnetic tetragonal state. Substitution of 18O for 16O in the separating layer of the type-1111 material has little effect on the magnetic or superconducting transition temperatures, indicating that magnetism and superconductivity occur primarily in the FeAs planes and are relatively independent of the separation layers.
For a simple isotropic BCS superconductor (which the arsenides are not) the maximum α is 0.5 when summed over all atoms in the compound. The large iron isotope effect found by Liu et al.1 for the superconducting transition would, at first glance, seem to indicate that phonons are not only important but are the dominant mediator for superconductivity. However, the superconducting transition temperatures determined using calculated values of the electron–phonon coupling constant and the phonon spectra, are less than 0.1 K. The calculated transition temperatures are so low that it is unlikely that phonons alone mediate the pairing. Even though the long-range magnetic order is removed by doping, short-range spin fluctuations will still be present. The coupling of these spin fluctuations to electrons is thought to be a possible mediator in the copper oxides and now in the iron arsenides as well. That both TSDW and Tc are affected by the iron mass (and with roughly the same sign and magnitude) indicates that they are somehow connected.
The glue for superconductivity in the iron arsenides could be magnetism (that is, the spin fluctuations) or possibly a combination of phonons and spin fluctuations. Although this result is suggestive, the interpretation of the IEC is not straightforward in these complex systems. The copper oxides show almost no oxygen IEC at optimal doping (maximum Tc) but show a value much larger than 0.5 in underdoped, low-Tc material. More work — both experimental, by measuring the IEC as a function of doping, and theoretical, by understanding how magnetism affects the superconducting IEC — will be required to fully understand these data. In the meantime, the results of Liu et al.1 underscore the fact that the excitement generated by the new iron arsenide superconductors is just beginning.
About this article
Physical Review B (2010)