Strain-engineering Mott-insulating La2CuO4

Abstract

The transition temperature T_c of unconventional superconductivity is often tunable. For a monolayer of FeSe, for example, the sweet spot is uniquely bound to titanium-oxide substrates. By contrast for La_2-xSr_xCuO₄ thin films, such substrates are sub-optimal and the highest T_c is instead obtained using LaSrAlO₄. An outstanding challenge is thus to understand the optimal conditions for superconductivity in thin films: which microscopic parameters drive the change in T_c and how can we tune them? Here we demonstrate, by a combination of x-ray absorption and resonant inelastic x-ray scattering spectroscopy, how the Coulomb and magnetic-exchange interaction of La₂CuO₄ thin films can be enhanced by compressive strain. Our experiments and theoretical calculations establish that the substrate producing the largest T_c under doping also generates the largest nearest neighbour hopping integral, Coulomb and magnetic-exchange interaction. We hence suggest optimising the parent Mott state as a strategy for enhancing the superconducting transition temperature in cuprates.

Fig. 1

Strain-dependent XAS and RIXS spectra of La₂CuO₄ films. a Normal incidence (δ = 25°) XAS spectra recorded around the copper L₃-edge on La₂CuO₄ films on different substrates as indicated. b–f display grazing incident copper L₃-edge RIXS spectra. In b, c, the dd excitations are shown for different momenta as indicated. For a–c, solid (dashed) lines indicate use of π (σ) polarised incident light. d displays the “centre of mass” of the dd excitations vs. strain ε, for samples as indicated. Each (thin film) point is an average “centre of mass” value of all the spectra in b, c. The bulk La₂CuO₄ value is extracted from ref. . e, f present the low-energy part of RIXS spectra (circular points) with four-component (grey lines) line-shape fits (see text). Notice that the different film systems have, naturally, different elastic components. For visibility all curves in a–e have been given an arbitrary vertical shift. g, h illustrate schematically the scattering geometry. Source data are provided as a Source Data file

Fig. 2

Magnon dispersions of La₂CuO₄ thin films. a, b display raw RIXS spectra recorded on the LCO/STO thin film system, along the antinodal [1 0] and nodal [1 1] directions, respectively. Red curves represent the data close to the antiferromagnetic zone boundary (AFZB) as shown in the inset in c. Solid lines are fits to the data (see text for detailed description). Notice that elastic scattering is, as expected, enhanced as the specular condition (0,0) is approached. In (c) magnon dispersions of LCO/LSAO and LCO/STO, extracted from fits to raw spectra as in a, b, along three different momentum trajectories (see solid lines in the inset) are presented. Solid lines through the data points are obtained from two-dimensional fits using Hubbard model (see the main text). Error bars are three times the standard deviations extracted from the fits. In (c) Q₁ takes different values for each compound due to slightly different incident energies and in-plane lattice parameters, resulting in 0.4437 (0.4611) for LCO/LSAO (LCO/STO). Source data are provided as a Source Data file

Fig. 3

Cuprate energy scales vs. strain. In (a) XAS at Cu L₃-edge resonances (left) and the theoretical results for Coulomb interaction U (right) are presented. Experimental and theoretical derived hopping parameters t are presented in b, as indicated. Notice that t is not scaling with the copper-oxygen bond length r^−α with α = 6–7 as sometimes assumed. J_eff as a function of ε = (a − a₀)/a₀ (where a₀ is the in-plane bulk lattice parameter) is presented in c for both theoretical and experimental results. Zone-boundary dispersion E_ZB, extracted from the Hubbard model, for experimental and theoretical parameters, are presented in the right inset of c as a function of strain ε. Superconducting transition temperature T_c as a function of out-of-plane lattice constant c—for optimally doped LSCO thin films (see Supplementary Table 1)—is presented in the left inset in c. Colour code for the data points in the figure refers to the one shown in c. The error bars for the experimental data are standard deviations extracted from the fits. The theoretical value corresponding to ε ≈ 0.01 (gray symbols) is an artificial sample as described in Table 1. Source data are provided as a Source Data file