Crystalline materials are produced in complex high-temperature processes involving a large variety of physical phenomena from heat transfer to fluid dynamics. We are working on a new generation of multiphysical models for these processes within the NEMOCRYS project funded by a Starting Grant from the European Research Council (ERC). The expected paradigm change in the way how we observe, describe and develop crystal growth processes will minimize the necessary experimental cycles and open new horizons for scientific analysis as well as for smart process control using artificial intelligence and other modern technologies.

We are developing a series of unique crystal growth setups for model materials to enable low working temperatures, relaxed vacuum-sealing requirements and hence an easy experimental access for various measurement techniques. The obtained in-situ measurement data are used to develop and validate multiphysical models, which are then applied for various crystal growth processes both on research and industrial scales. Furthermore, model experiments are employed to optimize the techniques for in-situ observation as well as to build a new basis for model benchmarking and for big data concepts in crystal growth.

Artificial intelligence (AI) based on artificial neural networks (ANN) can significantly reduce costs and shorten time for development of novel crystalline materials and crystal growth processes. The main reason for using ANNs is to quickly extract patterns and relationships in crystal growth data sets that are too complex to be noticed by other numerical techniques. The fast forecasting is crucial for the process control where in situ measurements of process parameters are not feasible.

The IKZ used static ANNs for parameter optimization in the magnetic driven growth of VGF-GaAs and DS-Si. A recent research topic was the application of dynamic neural networks for real-time prediction in the transient VGF-GaAs crystal growth process. For example, temperature distributions in the furnace as well as the position of the crystallization front during the growth can be quickly predicted. The AI application to other crystal growth topics is currently in progress. 

Crystal Growth is a kinetic process where atoms are integrated in a crystal lattice. This takes place for growth of bulk crystals and for deposition of crystalline layers. Understanding the kinetic processes is of great importance for the structure and the properties of the material as well as for the process engineering. Numerical calculations provide a deeper knowledge in concjunction with specific characterization methods such as e.g.  transmission electron microscopy (TEM) or atomic force microscopy (AFM). 

We develop a general kinetic Monte Carlo (KMC) tool using the programming language julia. The energy barriers for the various kinetic processes could be computed by potential models (Tersoff, ACE etc.) or via the bond counting method. They might be also taken directly from ab initio calculations for certain configuration.

This enables us to investigate the kinetics of different systems by atomistic computations using a unique framework. Currently, we investigate the epitaxial growth of AlN and AlGaN on AlN(0001), the homoepitaxy of Ga₂O₃ and the solution growth of Herbertsmithite.

In the research group Optimization and process control we specialize in the innovative application of mathematical optimization (MO) and advanced artificial intelligence methods (AI) to control and enhance crystal growth processes. Our vision is to redefine the methodologies used in crystal growth through the integration of state-of-the-art MO and AI technologies. We aim to develop control strategies that significantly enhance the efficiency of the growth process and quality of crystals.

Our main objective is tosupport equipment and process design by advanced modelling techniques. Specifically, we aim to:

• Innovative Techniques: Develop cutting-edge techniques that integrate physics-informed models with machine learning.
• Enhance Process Control: Improve control over crystal growth processes through innovative MO and AI applications.
• Quality and Efficiency: Optimize the growth conditions to maximize quality and operational efficiency.

Our group aims to advance the field of material science by fostering the development of superior crystal growth technologies. This work supports various industrial applications, contributing to advancements in electronics and renewable energy sectors.

We invite academic and industrial collaborators to join us in pushing the boundaries of materials science through advanced research.

Advanced MO and AI Methods
Our research leverages sophisticated MO and AI to develop control systems that optimize the parameters influencing crystal growth. These methods enable us to simulate and predict crystal growth behaviors with high accuracy, leading to improved outcomes.

Simulation and Optimal Control
By applying optimal control theories and simulation tools, we enhance the precision with which crystal environments are managed. Our goal is to consistently produce high-quality crystals by fine-tuning the conditions under which they grow.

Hybrid Modeling
We focus on creating hybrid models that integrate empirical and simulation data with AI-driven insights. These models are critical in developing robust predictive tools for crystal growth processes, ensuring better control and optimization strategies.

The principal goal of the ‘Solid State Theory’ group is to build ab initio quantitative modelling for the process occurring at surfaces and interfaces from the atomistic scale to the macroscopic scale and vice versa. The processes occurring at surfaces play a critical role in the growth, manufacture and performance of advanced materials, e.g., semiconductor thin film crystal growth. Such systems are controlled by atomistic processes, growth mechanisms, and their coupling involving a wide range of length and time scales that are difficult to probe by experiment alone. The ‘Solid State Theory’ group is focused on the development of ab initio adaptive multiscale modelling to adequately treat the dynamic nature of the processes in the semiconductor thin film crystal growth including 2D materials.

Method development

We are developing ab initio statistical mechanic methods to simulate the surface systems at scales relevant to real applications. In particular, we have developed a first-principles theory of replica-exchange grand-canonical molecular dynamics (REGC) that enables an accurate determination of surface phase diagram and dynamical description of surface restructuring under operation conditions. We are integrating the REGC method into the on-the-fly machine learning accelerated molecular dynamics to identify the (meta)stable phases and the elementary processes in the surface system under the relevant conditions. We are coupling the statistical learning techniques and the output of the REGC method to probe the surface kinetics, which will be incorporated in  macroscopic transport models to provide a detailed description of the local environment in  the reactor.

2D Materials and their heterostructures device

Large area continuous 2D crystals with uniform properties are of great importance for devices with reproducible performance. For practical applications of 2D materials, especially in electronics and optoelectronics, high quality and large area materials with controlled properties are prerequisites. This fact calls for a fundamental understanding of the material growth mechanism of 2D materials, i.e., how parameters such as the precursor, substrate, pressure, and temperature affect mass and heat transport, interface reactions and consequently the stoichiometric and atomistic arrangement of the 2D materials under the growth conditions. We ‘Solid State Theory’ group are focusing on study the atomistic mechanism of the 2D material in the CVD-based vapor phase growth methods, which provides strong support for the experimental nanostructure group. Furthermore, the predictive power could serve as guidelines to develop advanced growth method and to synthesize the new 2D materials and the 2D heterostructure in a controlled growth manner.

Semiconductor oxides thin film crystal growth

We are applying our in-house developed statistics mechanic methods to study the thermodynamics and kinetic of the oxides semiconductor surfaces under growth conditions. The goal is to build an intricate and dynamic relationship between structure-determined performance and growth environment for oxides semiconductors. By closely collaborating with the experimental groups in IKZ and industrial partners, we aim to accelerate the discovery of the next-generation oxides semiconductors and lead to a paradigm shift in the rational design of functional material for technological innovation.