Machine Learning Tools for the Measurements and Analysis of Quantum and Molecular Devices

A deep understanding of the electrical properties of quantum materials and molecular junctions remains very challenging and is currently limited to a small selection of non-reproducible devices which show large device-to-device fluctuations. Understanding the details of every single device remains difficult, time consuming and impossible to scale up. Fast screening of different materials including molecules, as well as precise understanding and control of their electronic properties will open the way to a large number of applications including quantum computing, sensing, etc. However, one of the main challenges of quantum hardware is the poor understanding and control of the properties of the material and its interfaces/surface.<br/><br/>In this talk, I will share our results on using benchmark systems and large dataset combined with machine learning tools to gain more insights into the understanding of transport properties in molecular devices. By developing high-throughput platforms, we could optimize the fabrication of graphene nanogaps to contact molecules and achieve the first stable and reproducible graphene molecular junction by decoupling the electronic properties of the junction from its mechanical anchoring and stability [1]. Combining large statistics allowed by mechanically controlled break junction and clustering techniques [2,3, we could also unravel several electronic paths through porphyrin molecules [4] as well as benchmark transport properties of alkane molecules [5]. I will conclude the talk with advance on the fabrication and measurements of carbon nanotube based spin qubits.<br/><br/>[1] <b>M. El Abbassi</b>, S. Sangtarash , X. Liu , M. Perrin , H. Sadeghi , O. Braun , H. van der Zant , S. Yitzchaik , S. Decurtins , S. L. Liu , C. Lambert and M. Calame. <i><u>Robust graphene-based molecular devices</u></i>. Nature Nanotechnology, 14, 957-967, 2019.<br/>[2]<b> M. El Abbassi</b>, Jan Overbeck, Oliver Braun, Michel Calame, Herre S.J. van der Zant, Mickael L. Perrin. <u>Benchmark and application of unsupervised classification approaches for univariate data</u><i><u>.</u></i><i>. </i><b>Communications Physics, 4, 50(2021).</b><br/>[3] Damien Cabosart, <b>Maria El Abbassi</b>, Davide Stefani, Riccardo Frisenda, Michel Calame, Herre SJ van der Zant, Mickael L Perrin. <i>A reference-free clustering method for the analysis of molecular break-junction measurements</i><i>. </i>Applied Physic Letters 114, 143102, 2019.<br/>[4] <b>M. El Abbassi</b>, P. Zwick, A. Rates, D. Stefani, A. Prescimone, M. Mayor, H. S. J. van der Zant and D. Dulic. <i><u>Unravelling the conductance path through single-porphyrin junctions</u></i><u>.</u> Chemical Science, 10, 8299-8305, 2019<br/>[5] Frederik H. van Veen, Luca Ornago, Herre S.J. van der Zant, <b>Maria El Abbassi</b>. <i><u>Benchmark study of alkane molecular chains</u></i>, just accepted to Journal of Chemistry C.