Research

Millions of calculated materials science data are now available to the scientific community (see for example the NOMAD project). Our research is focused on developing and implementing scalable and efficient computational methods to automatically extract knowledge from materials science data.

The starting point of our research are state-of-the-art data science techniques, such as for example convolutional and siamese neural networks, kernel methods, hierarchical clustering algorithms, and various dimensionality reduction methods. On top of this, we integrate our physical insight and domain knowledge in both descriptor identification (how the system is represented) and modeling. We strongly believe that the application of data science to materials should not only lead to transferable models with excellent performance, but more importantly generate value through real physical and chemical insight.

More specifically, we work in the following two areas:

  • Materials similarities: We develop methods to assess similarities and to build similarity maps between materials, these similarities being based on either structural, mechanical or chemical properties. These materials maps would reveal which regions of this high-dimensional space have not been explored yet, but may contain novel materials with unusual properties.
  • Crystal-structure classification: We use low-dimensional representations of physical systems (descriptors) and supervised learning techniques – in particular neural networks and kernel methods – to automatically classify crystal structures.

We are making the computational tools that stem from our research available to the scientific community with both easy-the-use and more advanced tutorials in the context of the NOMAD Analytics Toolkit.

Interests

  • Convolutional neural networks
  • Neural Networks interpretation
  • Non-linear Dimensionality Reduction
  • Big-data analytics

Current Research Project

  • Insightful classification of crystal structures using deep learning

    Interpretable deep learning solves the complicated task of structure recognition in nanomaterials.

    Computational methods that extract knowledge from materials science data are critical for enabling the data-driven discovery of novel nanomaterials for technological applications.

    A reliable identification of the crystal type is a crucial first step for materials characterization and analytics. Current methods require a user-specified threshold, and are unable to identify the correct crystal type for defective structures.

    Here, we introduce a new machine-learning-based approach to automatically classify nanomaterials by their crystal structure. First, we represent materials by a diffraction image, and then construct a deep-learning neural-network model for classification. Our approach is able to correctly classify a dataset comprising more than 80,000 structures, including heavily defective ones.

    The internal operations of the neural network are unraveled through attentive response maps, demonstrating that it uses the same landmarks a materials scientist would use, although never explicitly instructed to do so.

    Related publications:

    1. A. Ziletti, D. Kumar, M. Scheffler, and L.M. Ghiringhelli, Insightful classification of crystal structures using deep learning, Nature Communications 9, 2775 (2018) [pdf][link to journal]

Previous Research Projects

  • Phosphorene oxidation and phosphorene oxides

    Computational prediction and experimental realization of a new class of ultrathin (one-atom thick) materials

    After its isolation in early 2014, phosphorene – a single layer of black phosphorus – has been propelled to the limelight because of its exceptional electrical properties, which make it probably the most promizing material for the next generation of computers. If left in air, however, phosphorene degrades quickly, loosing its valuable properties.

    In this project, we theoretically study for the first time the oxidation of phosphorene, showing that oxygen atoms can be readily incorporated in the material, leading in some cases to drastic changes in the phosphorene electric properties. This is a curse because oxidation leads to the degradation of the electrical properties, but also a blessing since simple air-exposure will lead – according to our prediciton – to the creation of new low-dimensional phosphorus-oxygen nanomaterials.

    These predictions are later experimentally confirmed by high-resolution surface sensitive photoelectron spectroscopy measurements. Comparison with our calculations shows that a single layer of phosphorus oxide (P4O10) is formed at the surface of black phosphorus after prolonged air exposure. This oxide not only constitutes a natural protective layer for the black phosphorus layers beneath, but it also possesses unique optical and fluorescence properties. Finally, we also show that such these new materials (phosphorene oxides) can be created on-demand by laser illumination, and can even be used as a miniaturized toxic gas monitor.

    Related publications:

    1. A. Ziletti, A. Carvalho, D.K. Campbell, D.F. Coker, and A.H. Castro Neto, Oxygen Defects in Phosphorene, Physical Review Letters 114, 046801 (2015) [pdf] [link to journal]
    2. R.A. Doganov, E.C.T. O’Farrell, S.P. Koenig, Y. Yeo, A. Ziletti, A. Carvalho, D.K. Campbell, D.F. Coker, K. Watanabe, T. Taniguchi, A.H.Castro Neto, and B. Ozyilmaz, Transport properties of pristine few-layer black phosphorus by van der Waals passivation in an inert atmosphere, Nature Communications 6, 6647 (2015) [pdf] [link to journal]
    3. A. Ziletti, A. Carvalho, P.E. Trevisanutto, D.K. Campbell, D.F. Coker, and A.H. Castro Neto, Phosphorene oxides: Bandgap engineering of phosphorene by oxidation, Physical Review B 91, 085407 (2015)[pdf] [link to journal]
    4. A. Ziletti, S.M. Huang, D.F. Coker, and H. Lin, Van Hove singularity and ferromagnetic instability in phosphorene, Physical Review B 92, 085423 (2015) [pdf] [link to journal]
    5. J. Lu, J. Wu, A. Carvalho, A. Ziletti, H. Liu, J.Y. Tan, A.H. Castro Neto, B. Ozyilmaz, and C.H. Sow, Bandgap Engineering of Phosphorene by Laser Oxidation towards Functional 2D Materials, ACS Nano 9(10), 10411 (2015) [pdf] [link to journal]
    6. M.T. Edmonds, A. Tadich, A. Carvalho, A. Ziletti, K.M. O’Donnell, S.P. Koenig, D.F. Coker, B. Ozyilmaz, A.H Castro Neto, and M.S. Fuhrer, Creating a Stable Oxide at the Surface of Black Phosphorus, ACS Applied Materials & Interfaces 7, 14557 (2015) [pdf] [link to journal]
  • Coherent transport in quantum circuits

    Analytical and numerical study of electon trasport in nanostructures

    The miniaturization of electronic components follows an impressive pace, and very soon electronic circuits will have to operate in a new regime in which a quantum mechanical treatment is needed.
    In this project, we consider a multibranch quantum circuit which is a paradigmatic system in quantum information. We study both the closed and the open system, where in the latter the quantum circuit is coupled to the environment through a non-Hermitian Hamiltonian approach. We find that long-lived localized states are formed in the middle of the circuit, fact that can be important for the fabrication of new kinds of electronic nanostructures, waveguides, antennas, and memory devices.

    Related publication:

    1. A. Ziletti, F. Borgonovi, G. L. Celardo, F. M. Izrailev, L. Kaplan, and V. G. Zelevinsky,
      Coherent transport in multibranch quantum circuits, Physical Review B 85, 052201 (2012) [pdf] [link to journal]