CAREER: Emergence of in-liquid structures in metallic alloys by nucleation and growth

结晶是一种众所周知的现象，其中固体区域开始在非常小的长度尺度上在液体内形成。这种现象是自然现象和工程应用的基础，包括冰的形成、天然气管道中的烃笼形物、骨形成期间磷酸钙的生物矿化、用于药物设计和生产的分子晶体的合成以及金属合金的固化以实现期望的机械性能。固体区域的形成高度依赖于下面的液相结构。在液体中，一些原子和/或分子可以自组织成几何或非几何结构，并在很短的时间内存在。这些结构被称为涌现结构，有时可以将自己转化为新的纳米级结构，从而触发结晶过程。根据材料的不同，转化可能涉及多个步骤，甚至可以是级联过程，其中一种结构导致另一种结构，所有这些都发生在液相中。 对液体中出现的结构和多步结晶过程的理解是至关重要的，因为每一步的结构都可能包含有用的信息，并揭示可以用于所需应用的特性。不幸的是，液体结构的物理询问通常基于昂贵的定制仪器，即，原位X射线同步加速器衍射往往受到时间和空间分辨率限制。另一方面，原子模拟是，在原则上，能够提供一个详细的，空间和时间上解决，表征在液体紧急结构到晶体的转换机制。然而，这些模拟可能非常耗时，因为它们需要在多次迭代中计算大量原子之间的量子力学相互作用。现代机器学习和人工智能方法有望规避原子模拟目前面临的限制。PI的团队将联合收割机原子模拟与机器学习和人工智能方法相结合，以获得对液体中涌现结构及其对结晶的影响的前所未有的见解。该项目将整合研究和教育，为来自代表性不足的社区的本科生和研究生建立加入新墨西哥州Tech的途径。这一努力认识到，新墨西哥州是一个西班牙裔人口占多数的州，也是一个重要的美洲原住民社区，目前在获得STEM项目方面面临限制。为此，PI将开展两个项目：（1）为当地高中的10- 11年级学生和教师开发为期两个月的夏令营，名为Camp PyMatter，这将使用Python编程语言介绍材料科学的基础知识，以及（2）通过开发交互式软件，与纳瓦霍技术大学的工程学院和数学技术学院进行外联&。大学教师的合作，并与合金制造技术和计算材料科学为基础的主题吸引他们的学生。技术总结最近的研究表明，复杂的结构，出现在液态催化结晶通过多步成核过程。令人惊讶的是，这些结构与最终的平衡固体几乎没有相似之处。然而，现存的经典成核和生长理论假设通过单步过程从液相到平衡固体的直接转化，并且不考虑这样的结构。该项目旨在通过开发一个集成的，基于机制的建模框架来克服这一基本限制，以预测多步成核和生长途径，并量化其能量学和动力学，以解决各种结晶问题。PI的研究团队将通过研究模型金属合金来实现以下三个研究目标：（1）开发一种稳健的方法来检测液相内的涌现结构并将它们与能量结构景观相关联，（2）利用该景观来量化与多步成核过程相关联的活化能和成核速率，以及（3）利用能量结构景观来量化固体的生长动力学。原子模拟将用于检测结构并确定其能量，并通过使用无监督，监督和生成式机器学习方法建立结构-热力学关系。该项目将直接影响用于承载应用的新型金属合金的发现。它将指导选择适当的合金元素，有效地影响形核能量和速度从液体熔体凝固过程中。通过对形核过程的控制，可以精确地控制凝固组织的晶粒尺寸分布和机械性能。本研究中使用的框架是强大的，适应性强，可扩展的，因为它允许检查新元素添加对当前合金凝固的影响，并且至关重要的是，为发现具有变革性机械性能的下一代材料铺平了道路。高熵合金的开发就是一个例子，它强调了由于其多元素环境而产生的复杂结晶机制的作用。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估而被认为值得支持。

NONTECHNICAL SUMMARYCrystallization is a well-known phenomenon where solid regions start forming within a liquid on very small length scales. This phenomenon is foundational to natural phenomena and engineering applications, including ice formation, hydrocarbon clathrates in natural gas pipelines, bio-mineralization of calcium phosphate during bone formation, the synthesis of molecular crystals for drug design and production, and solidification of metallic alloys to achieve desirable mechanical properties. The formation of solid regions is highly dependent on the underlying liquid phase structure. In the liquid, a few atoms and/or molecules can self-organize themselves into geometric or non-geometric structures that exist for a very short time. These structures, which are called emergent structures, can sometimes convert themselves into new nanoscale structures that trigger the crystallization process. Depending on the material, the conversion may involve multiple steps or can even be a cascading process, where one structure leads to another structure, all happening within the liquid phase. An understanding of the in-liquid emergent structures and the multistep crystallization process is critical because the structure at each step may hold useful information and reveal properties that can be leveraged for desired applications. Unfortunately, physical interrogation of the liquid structure is often based on expensive custom-built instrumentations, i.e., in situ X-ray synchrotron diffraction, that tend to be limited by temporal and spatial resolution. On the other hand, atomistic simulations are, in principle, able to provide a detailed, spatially and temporally resolved, characterization of the in-liquid emergent structure-to-crystal conversion mechanisms. Yet, these simulations can be prohibitively time-consuming because they require computing quantum-mechanical interactions between a large number of atoms over multiple iterations. Modern machine-learning and artificial intelligence approaches promise to circumvent limitations currently faced by atomistic simulations. The team of the PI will combine atomistic simulations with machine-learning and artificial intelligence approaches to gain unprecedented insights into in-liquid emergent structures and their influence on crystallization.This project will integrate research and education to establish a pathway of undergraduate and graduate students from underrepresented communities to join New Mexico Tech. This effort recognizes that New Mexico is a state with a Hispanic majority population and a significant Native American community that currently face limitations in gaining access to STEM programs. To this end, the PI will pursue two projects: (1) the development of two month-long summer camps, named Camp PyMatter, for 10-11th grade students and teachers from local high schools, which will introduce the basics of Materials Science using the Python programming language, and (2) outreach to Navajo Technological University’s School of Engineering, Math & Technology by developing inter-university faculty collaborations and engaging their students with alloy fabrication techniques and computational Materials Science-based topics.TECHNICAL SUMMARYRecent studies have revealed that intricate structures that emerge within the liquid state catalyze crystallization via multi-step nucleation processes. These structures bear surprisingly little resemblance to the final equilibrium solid. However, extant classical nucleation and growth theories assume a direct transformation from the liquid phase to equilibrium solid via a single-step process and do not account for such structures. The project seeks to overcome this fundamental limitation by developing a thermodynamically integrated, mechanism-based modeling framework to predict muti-step nucleation and growth pathways and quantify their energetics and kinetics for a wide-range of crystallization problems. The PI’s research team will pursue the following three research objectives by studying model metallic alloys: (1) develop a robust methodology to detect emergent structures within a liquid phase and correlate them with energy-structure landscape, (2) utilize that landscape to quantify activation energy and nucleation rates associated with multi-step nucleation processes, and (3) leverage energy-structure landscape to quantify growth kinetics of a solids. Atomistic simulations will be employed to detect structures and determine their energies, and establish structure-thermodynamics relationship by using unsupervised, supervised, and generative machine learning methods. This project will directly impact the discovery of novel metallic alloys used in load-bearing applications. It will guide the selection of appropriate alloying elements that effectively influence the nucleation energetics and rates during solidification from a liquid melt. By exerting control over the nucleation process, one can precisely manipulate the grainsize distribution and mechanical properties of the solidified microstructure. The framework used in this study is robust, adaptable, and scalable, as it allows for the examination of the effects of novel elemental additions on solidification of current alloys and, crucially, pave way for discovering next-generation materials with transformative mechanical properties. One such example is the development of high entropy alloys, which emphasize the role of complex crystallization mechanisms due to their multi-element environment.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.