Machine learning-guided substrate temperature optimization for anisotropic Fe(Se, Te) superconducting thin films

Abstract

The comprehension of superconductivity in Fe(Se, Te) is fundamentally tied to the material's anisotropy and the associated physical properties. In our research, we have prepared a series of high-quality Fe(Se, Te) films on a CaF₂ substrate using the pulsed laser deposition technique. Throughout our investigation, we focused on how the structure and superconducting characteristics evolve in response to varying growth temperatures. To quantitatively identify the optimal substrate temperature window for high-quality film synthesis, machine learning models were integrated, correlating growth parameters with superconducting performance metrics. Under identical magnetic field conditions, we found that the flux pinning energy in the ab-plane orientation surpasses that observed in the c-axis orientation. This observation clearly points to an anisotropic response within the material. The transition observed may suggest a shift in the predominant flux pinning mechanism, transitioning from point pinning to surface pinning, which highlights the complex nature of the superconducting vortex dynamics in these films. Our observations revealed a subtle yet significant temperature dependence of the anisotropic parameter γ, particularly near the critical temperature (T_c) for films grown at 300 °C. Machine learning predictions further refined the operational temperature window to 280–320 °C, aligning with experimental validation of peak superconducting performance at 300 °C. This behavior underscores the intricate relationships between growth conditions, temperature, and the resulting superconducting properties in Fe(Se, Te) films.