Characterisation of Crystalline Materials through Imaging, Image Processing and Machine Learning for 3D Shape Description

有机材料的晶体生长在特种和精细化学工业中具有重要意义。这反映了其在材料纯化中的实用性及其在制备具有最佳产品性能所需的明确定义的晶体尺寸、形状和多晶型形式的广泛化合物中的用途。后者很重要，例如，确保可重现的溶解和稳定性行为，以保持配制产品中成分的安全性和有效性。有机精细化学品的固有复杂性直接影响其物理化学颗粒性质，特别是其在具有各向异性形态和表面性质的低对称晶体结构中的结晶。这些性质的变化或可变性可影响材料的下游性能，例如，生物利用度、粉末处理、稳定性和可制造性。已知形成固溶体的杂质会影响形态（形状以及表面性质）。目前的颗粒尺寸测量在形状表征方面可能过于简单化，主要集中在球形颗粒上。这种方法不能反映精细化学品中重要的晶体形态，其中不同的晶面可能具有不同的表面化学，因此与其加工环境的分子间相互作用不同。目前，在能够将分子结构与颗粒形式的相同材料的性能相关联的能力方面存在关键差距。这种知识差距导致人们对融合分子晶体学数据、模拟特性和人工智能（AI）方法的兴趣越来越大。该项目的目的是通过应用数字AI技术来开发基于形态学的形状描述符，以精确表征3D中的晶体颗粒，从而满足上述需求。为此，机器学习将被应用于将来自过程显微镜的图像映射到3D晶体形状和功能特性的描述。该项目将有助于使有机晶体材料的设计更严格的粒度/形状规格，具有更高的一致性和更少的可变性。该项目将探索工艺杂质的影响，并旨在模拟固溶体对晶体生长和表面特性的影响。学生将结晶技术与人工智能/机器学习相结合，用于结晶工艺工程，重点是3D晶体表征。该项目的目标包括1。密集的文献回顾2.开发和调试一个新的原位细胞的动态表征晶体形态和生长3。形态学表征和晶体生长动力学的实验研究，收集在不同工艺条件下制备的晶体图像4.使用分子和晶体学建模来分析原位显微镜数据，并将其与AI/机器学习技术相结合5。将方法扩展到在线成像，用于监测批量处理期间晶体群体生长的动态。该项目围绕粒子技术和人工智能/机器学习领域展开，监督团队中有这两个领域的专家。该项目将是实验密集型的，利用利兹的结晶实验室，这些实验室装备齐全，原位显微镜、结晶系统（1 mL至20 L）、固态表征（Keyence数字显微镜、IGC/GC、DSC、TGA、FTIR、UV-vis、Morphologi G3等）。该项目的工业合作伙伴先正达的支持将为工业案例研究提供安置机会。

The crystal growth of organic materials is of significant importance within the speciality and fine chemical industries. This reflects its utility in materials purification and its use in preparing a wide range of compounds which have the well-defined crystal size, shape and polymorphic form needed for optimal product performance. The latter is important e.g., in ensuring the reproducible dissolution and stability behaviour needed to maintain the safety and efficacy of ingredients within formulated products. The inherent complexity of organic fine chemicals directly impacts on their physical chemical particulate properties, notably their crystallisation in low symmetry crystallographic structures with anisotropic morphologies and surface properties. Changes to, or variability in, these properties can affect downstream performance of the material, e.g., bioavailability, powder handling, stability and manufacturability. Impurities, that form solid solutions are known to influence morphology (shape as well as surface properties). Current particle sizing measurements can be over-simplistic in terms of shape characterisation being focussed mostly on spherical particles. Such methods do not reflect crystal morphologies important in fine chemicals where different crystal faces can have different surface chemistry and hence different intermolecular interactions with their processing environments. Currently, there is a critical gap in capability to be able to relate molecular structure to performance of the same material in particulate form. This knowledge gap has led to increasing interest in fusing molecular crystallographic data, simulated properties and artificial intelligence (AI) approaches. The aim of this project is to address the above need by applying digital AI-enabled technology to develop morphologically-based shape descriptors for precisely characterising crystalline particulates in 3D. To do this machine learning will be applied to map the images from in-process microscopy to a description of 3D crystal shape and functional properties. The project will help enable the design of organic crystalline materials to a much tighter particle size/shape specification with more consistency and less variability. The project will explore the impact of process impurities and aim to model the impact of solid solutions on crystal growth and surface properties. The student will integrate the crystallisation technology with AI/machine learning for crystallisation process engineering with a focus on 3D crystal characterisation. The project aims encompass 1. Intensive literature review 2. Development and commissioning of a new in-situ cell for the dynamic characterisation of crystal morphology and growth 3. Experimental studies of the morphological characterisation and crystal growth kinetics collecting crystal images prepared under varying process conditions 4. Using molecular and crystallographic modelling to characterise in-situ microscopy data integrating this with AI/machine learning techniques 5. Extending approaches to online imaging for monitoring the dynamics of the growth of a population of crystals during a batch processing. The project is centred around the areas of particle technology and AI/machine learning with experts within the supervisory team in both areas. This project will be experimentally intensive with making use of the crystallisation laboratories at Leeds which are well equipped with e.g., in-situ microscopes, crystallisation systems (1mL to 20L), solid-state characterisation (Keyence Digital Microscope, IGC/GC, DSC, TGA, FTIR, UV-vis, Morphologi G3 etc). Support by the project's industrial partner, Syngenta, will provide placement opportunities for industrial case studies.