Symposium DS03—Phonon Properties of Complex Materials—Challenges in Data Generation, Data Availability and Machine Learning Approaches

This symposium will broadly cover current and emerging data generation techniques and data driven analysis approaches to characterize phonons, the quantized vibrations of condensed matter systems. Phonons play an increasingly important role in information-processing applications, both directly and indirectly through interactions with other quasiparticles and energy carriers. A key focus of the symposium will remain on thermal properties of materials enabling such applications. Additionally, studies exploring co-optimization of properties of multiple carriers (e.g., electron and phonon) in a variety of materials, such as thermoelectrics, wide-bandgap semiconductors, and photovoltaics, will be of particular interest. The first part of the symposium will focus on emerging theoretical and experimental techniques to calculate/measure phononic properties of complex materials. Symposium contributions should address basic science issues or highlight exploration of unusual phenomena (e.g., glass like phonon transport in crystals and/or low-symmetry materials), and address challenges in understanding the corresponding physical mechanisms. Discussion of theoretical, computational or experimental characterization techniques, challenges in data generation and applicability of emerging materials for technologies are also welcomed. The second part will focus on machine learning (ML) approaches for phononic property prediction, that are of mutual interest of the broader materials informatics communities. ML-enabled design and discovery of new materials are increasingly being facilitated by large amounts of data available through databases, however, the availability of phonon properties data is limited. Discussion of development and application of physics-based ML models that can work with sparse data and provide consistent validation approaches are particularly of interest. Contributions discussing availability of data and methods to improve data sharing practices are also of interest.

• Emerging phonon dynamics in complex materials
• Advances in theoretical, computational and experimental phononic property characterization techniques
• Co-optimization of multiple carrier (e.g., phonon, electron) properties for thermoelectrics and other emerging technologies
• Machine learning studies for prediction of thermal properties of nanostructured materials
• Machine learning studies probing interaction of phonons with electrons and other quasiparticles
• Data driven studies for characterization of vibrational properties of complex materials
• Data mining of thermal imaging data
• Phononic property data generation and sharing practices
• Challenges in developing machine learning algorithms with limited training data
• Novel validation approaches to test machine learning model predictions

• Maria Chan (Argonne National Laboratory, USA)
• Stefano Curtarolo (Duke University, USA)
• Pierre Darancet (Argonne National Laboratory, USA)
• Geoffroy Hautier (Dartmouth College, USA)
• Run Hu (Huazhong University of Science & Technology, China)
• Shenghong Ju (Shanghai Jiao Tong University, China)
• Tengfei Luo (University of Notre Dame, USA)
• Apurva Mehta (SLAC National Accelerator Laboratory, USA)
• Jesús Carrete Montaña (TU Wien, Austria)
• Kristin Persson (University of California, Berkeley, USA)
• Xiulin Ruan (Purdue University, USA)
• Abhishek Singh (Indian Institute of Science, Bengaluru, India)
• Sebastian Volz (The University of Tokyo, Japan)
• Chris Wolverton (Northwestern University, USA)
• Yibin Xu (National Institute for Materials Science, Japan)