Hetero-aggregation of nanoparticles in advective environments

混合和聚集是制药和化工行业中的关键过程，过程控制对产品的特性有着显著的影响。如果使用多种材料，异质接触的数量，即不同组分的颗粒之间的接触数量，对产品的性能至关重要。很可能，该数量可以通过混合的顺序来控制，并且在宏观尺度（将组分聚集在一起）和微观尺度（导致碰撞和接触形成）下的混合的相对强度将对最终产品质量起决定性作用。虽然小颗粒的碰撞动力学，不受平流的影响是很好的理解，可以模拟体面的准确性，有没有经过验证的方法的统计描述的异质聚集的颗粒大小是远远不是单分散的，在颗粒分布的初始不均匀性有很大的影响，最终的总结构和质量。拟议的工作旨在解决与聚集的两组分系统的可变的初级颗粒尺寸范围从约5至200 nm的问题。将进行Langevin动力学（LD），以生成各种工艺条件下的聚集体及其粒度分布。LD为这种研究提供了一个合适的工具，因为它将跟踪每一个初级颗粒，并且不会对所产生的聚集体结构做出任何假设。模拟允许的异质聚集体，包括它们的质量（异质接触的数量）的表征和作为聚集体特性和流动条件的函数的碰撞频率的分析。传统的碰撞频率模型不太可能成立，取而代之的是建议使用深度学习方法来关闭种群平衡方程。

Mixing and aggregation are key processes in the pharmaceutical and chemical industries, and process control can markedly influence the products’ characteristics. If multiple materials are used, the number of hetero-contacts, i.e. the number of contacts between particles of the different components, are crucial for the product’s properties. It is likely that this number can be controlled by the sequence of mixing, and the relative strength of mixing at the macroscale (that brings the components together) and at the microscale (that leads to collision and contact formation) will be decisive for the final product quality. While collision dynamics for small particles that are not influence by advective flow are well understood and can be modelled with decent accuracy, there is no validated approach for a statistical description of hetero-aggregation where particle sizes are far from being monodisperse and where initial inhomogeneities in particle distribution have a large effect on the final aggregate structure and quality. The proposed work intends to address issues related to aggregation of a two-component system with variable primary particle sizes ranging from about 5 to 200 nm. Langevin dynamics (LD) will be conducted to generate aggregates and their size distributions for a variety of process conditions. LD provide a suitable tool for such study as every single primary particle will be tracked and no assumptions will be made on the resulting aggregate structure. The simulations allow for a characterization of the hetero-aggregates including their quality (number of hetero-contacts) and for an analysis of collision frequencies as function of the aggregate characteristics and flow conditions. Conventional models for collision frequencies are not likely to hold, and deep learning methods for closure of the population balance equation is suggested instead.