2-10. High-throughput Characterization based on Far-Field Optical Probe

X.-D. Xiang
Southern University of Science and Technology

Abstract: The systematic relationships between material composition, structure, synthesis parameter and property, generally presented as “phase diagrams”, are the core data of material science and engineering. Taking advantages of unprecedented computing power, material research is transforming to the 4th paradigm featuring intensive data and artificial intelligence (AI). AI provides a powerful tool to overcome human limitation in unravelling multi-dimensional, complex material data to quickly identify the relationship between application-oriented properties and corresponding parameters. The success of advanced AI methods, such as neural networks, relies on availability of intensive material data. However, the basic data of materials science is extremely scarce. 

        High-throughput computational simulation can produce massive material data, only if simulation models are validated with experimental results. High-throughput experimentation can generate large volumes of experimental data via combinatorial material synthesis and rapid characterization. However, owing to the lack of high-throughput characterization techniques, it is difficult to achieve experiment-validated and sophisticated material big data, which becomes the bottleneck of data-driven material research.

        Far-field technique played a critical role in realization of Moore's Law in IC, astrophysics and life sciences. Inspired by this, the talk will discuss the key to break the bottle-neck of materials data growth rate - through far-field optical methods to acquire physical properties of materials.
Keywords: material data science; high-throughput experiments; far-field optical characterization

    基于远场光探针的光、电、热、磁、力高通量表征

项晓东
南方科技大学

摘要: 材料成分，结构，工艺和性能之间的关联是材料科学的核心数据。在材料科学领域，人 工智能近些年开始展现出强大的潜在威力，其通过学习密集型材料数据，可分析总结出目标性 能与核心参数的关系，从而快速预测、优选出具有目标性能的新材料。先进人工智能方法，诸 如神经网络的成功，完全依赖于数据的数量和质量。根据生命科学、宇宙科学领域的经验，一 个领域的数据如果按照摩尔定律增长，人工智能就能在该领域取得成功。然而材料科学基础数据极度匮乏，很难支撑数据驱动模式的发展。

      伴随计算模拟能力的不断提升，基于高通量计算的材料数据获取越来越快。然而对计算数据的实验验证仍相当匮乏，只有将计算数据与实验数据相互验证，进一步修正计算模型，才能保证数据的准确性。同时，高通量实验可快速提供有价值的研究成果，直接加速材料的筛选和 优化。材料实验领域已发展多项成熟的高通量材料制备技术，但仍缺乏对材料性质参数进行快速测量的成熟高通量表征技术，难以实现海量实验数据的快速获取，成为材料数据科学发展的瓶颈。

      远场光刻技术使得集成电路芯片中的晶体管数量根据摩尔定律增长成为可能。光学探针技术和远场光学显微镜在基因测序技术的发展中也起到了同样的作用。受远场技术在集成电路和 基因测序中成功应用的启发，本报告将重点讨论高通量远场光探针实验室表征技术，该技术可 快速、完整、准确地表征材料结构、成分、价态、配位、光、电、热、磁、力性能数据，有望 打破材料数据增长瓶颈，实现材料数据量按照摩尔定律描述的速度增长，保证人工智能在材料 科学领域的成功应用，最终实现材料科学由“试错法”向“数据+人工智能”科学“第四范式” 的根本转变。

关键词:材料数据科学;高通量实验;远场光学表征技术