Superconductors under stress

Abstract

Locquet et al.¹ reported a doubling of the critical temperature (T _c) of the superconductor La_1.9Sr_0.1CuO₄, from 25 to 49 K, in compressively strained thin films on a SrLaAlO₄substrate. This led to speculation that values of T _cclose to or even beyond 200 K could be obtained without excessive external pressure, starting from an unstrained cuprate with high T _c(>100 K). The stress would be applied to the ab (CuO₂) plane of this system of tetragonal symmetry. Such expectations are theoretically unrealistic.

Main

To understand the phenomenon they observed, Locquet et al. cited data on uniaxial pressure gradients of T _con the same compound²,³. In particular, they inferred from the negative sign of the quantity dT _c/dɛ_c, where ɛ_cis the strain (minus the relative change in distances along the c -axis), that elongation (negative strain) of the c -axis increases the critical temperature, if atomic coordinates in the ab plane are kept constant.

Without questioning the validity of their experimental results, we disagree with Locquet et al. on two points. First, the higher the critical temperature in the unstrained state, the closer is the system with respect to optimum density in coordinate space, or to the optimum level of the Fermi energy. Quenching will then lead to a much weaker effect than at low T _c, and may even lower the critical temperature. This conclusion reflects apriori on results selecting the Hg-cuprate with the present record T _cof 134 K (three CuO₂layers per molecular unit).

Second, the cuprates La_21xM_xCuO₄(where M is Sr or Ba) are unsuited as a reliable source of microscopic information via stress-strain relations because of observed structural distortions, phase transitions, non-rigid tilting of CuO₆octahedra, and so on³, under external parameters and doping. As a result, conventional stress-strain relations, which are macroscopic and phenomenological, are not applicable to these systems on an atomic scale. This implies that the usual equalities (all derivatives are partial), dT _c/dɛ_i = ∑_j(dT _c/dP _j)C _ji, where ɛ_iis the strain along the i th axis and C _jiare elastic moduli, cannot provide the information desired. With i denoting the c -axis direction, the right-hand side is large and negative³,⁴ for La_21xSr_xCuO₄, but its indicated relation to a quantity which presupposes that all coordinates except those along the c -axis remain unchanged does not hold.

Locquet et al. incorrectly conclude that these negative values imply an increase of T _cbrought about solely by enlarging the c -axis. In particular, their criticism of the interlayer tunnelling model proposed by Anderson et al.⁴ is irrelevant, although this model has been criticized on other grounds⁵.