Discovering excitonic superconductors

在环境压力下发现室温超导性有可能彻底改变许多行业，并推动能源技术的重大进步。实现高温超导性的一种拟议机制涉及通过光诱导过程形成激子，这可以促进电子配对，并在比传统声子介导的超导体高得多的温度下实现超导电流。虽然先前的研究确定了碳层上的过渡金属二硫属化物（TMD）作为观察激子介导的超导性的有希望的候选者，但尚未观察到实验证实。在这项工作中，我们采用物理信息机器学习方法来识别，合成和测试碳层上的TMD作为潜在的激子超导体候选者。我们提出的框架涉及使用通过自动第一性原理计算获得的带隙和激子结合能等计算属性构建机器学习模型。此外，我们整合了实验获得的数据和材料数据库（如材料项目）中的现有信息，以增强机器学习模型的预测能力。在TMDs属性预测的指导下，我们进行合成实验，主要采用热化学方法合成TMDs纳米颗粒，随后用于涂覆碳纤维，或涂覆有碳本身。然后使用光学低温恒温器对合成的材料进行表征和超导性测试，使光穿透，同时将样品冷却到接近绝对零度的温度。所获得的数据随后用于改进机器学习模型的预测，合成和表征的循环继续进行，从而揭示新型TMD材料的特性。

The discovery of room temperature superconductivity at ambient pressure has the potential to revolutionize numerous industries and drive significant advancements in energy technologies. One of the proposed mechanisms for achieving high-temperature superconductivity involves the formation of excitons through light-induced processes, which can facilitate the pairing of electrons and enable superconducting currents at significantly higher temperatures than traditional phonon-mediated superconductors. While previous research identified transition metal dichalcogenides (TMDs) on carbonaceous layer as promising candidates for observing exciton-mediated superconductivity, experimental confirmation has yet to be observed. In this work, we employ a physics-informed machine learning approach to identify, synthesise and test TMDs on a carbonaceous layer as potential excitonic superconductor candidates. Our proposed framework involves construction of a machine learning model using computed properties such as band gap and exciton binding energies obtained through automated first-principles calculations. Additionally, we integrate experimentally acquired data and existing information available in materials databases, such as the Materials Project, to enhance the predictive capabilities of the machine learning model. Guided by the TMDs property predictions, we perform synthesis experiments mainly employing thermochemical methods to synthesise TMDs nanoparticles that are subsequently used to coat carbon fibres, or that are coated with carbon themselves. The synthesised materials are then characterized and tested for superconductivity using an optical cryostat, enabling light penetration while cooling the samples to near absolute zero temperatures. The acquired data are subsequently used to refine the machine learning model's predictions, and the cycle of synthesis and characterization continues, leading to uncovering properties of novel TMD materials.