(2cb) Development of an Integrated Multiscale Modeling, and Control Framework for Commercialization of Quantum Dot Manufacturing and Their Applications

The US Department of Energy has set high-priority targets to reduce the cost of photovoltaics from ~ 10 cents/kWh to half of that at 5 cents/kWh by 2030. To achieve this objective, large-scale production of highly efficient and cost-effective perovskite quantum dot (QDs)-based solar cells is necessary. However, there are major commercialization challenges associated with the manufacturing of QDs: (a) a lack of understanding of the crystallization kinetics of various QD systems, which hinders the predictive control of the QD size distribution; (b) a majority of existing QD synthesis techniques follow lab-scale batch protocols, and there are very few demonstrations of fast-scalable continuous production of QDs, which is essential for validating QD scale-up frameworks; and (c) lack of accurate models and adequate controllers for regulation of QD thin-film deposition (i.e., film thickness and roughness), which is an essential step in different QD-based optoelectronic applications (i.e., solar cells and high-resolution displays). All these challenges need to be addressed in an effective, and step-by-step manner to create a foundation for the commercialization of QDs in the near future. Thus, there is a need for an integrated end-to-end framework that addresses each of the above challenges by combining various appropriate modeling and control techniques.

To address these knowledge gaps, we (a) developed a combination of different models to describe the mechanism of QD crystal growth and enable fast-scalable manufacturing of QDs and the associated optoelectronic devices; (b) developed appropriate control systems to regulate various QD properties (i.e., size distribution) for application in various optoelectronic devices; and (c) integrated the above-developed models and controllers into a single fully integrated framework to address the problem of commercialization of QD-based devices.

First, a first-principled kinetic Monte Carlo (kMC) was developed and experimentally validated to describe the crystallization kinetics of QDs. Second, to resolve the various issues associated with the batch synthesis of QDs, continuous manufacturing of QDs using a plug flow crystallizer (PFC) was demonstrated using a high-fidelity (HF) multiscale modeling approach. Specifically, since the developed HF model was computationally expensive, a highly efficient Artificial Neural Networks (ANN) was utilized as a surrogate model and was incorporated into a data-driven optimal operation problem to regulate QD crystal size and distribution. Further, this approach was extended to two-phase slug-flow crystallizers (SFCs) by constructing a CFD-based multiscale model to account for the slug-to-slug variation and its effect on the QD size distribution. Third, modeling of process steps required for manufacturing solar cells and high-resolution displays was performed to address the commercialization challenges in the QD industry. Specifically, (a) a thin-film deposition framework comprising of macroscopic heat and mass transfer and microscopic discrete element method (DEM)-based aggregation model was developed for describing the spray coating of QDs; and (b) a kMC model was developed to model nanopatterning of QDs for application in high-resolution displays. Lastly, it is important to note that all the above-developed models were experimentally validated using appropriate experimental observations.

Overall, the proposed work addresses three major challenges in the QD field such as the control of QD kinetics, continuous production of QDs, and designing manufacturing processes for fast scale-up of QD-based devices. This was executed by developing various experimentally validated models (i.e., the kMC model for QD crystallization, CFD-based model for SFCs, a DEM-based spray coating model, and others). Also, model predictive control systems that utilized the above-develop models were designed to regulate the various QD properties (i.e., size and distribution, and roughness of thin films). Finally, the abovementioned models and controllers were combined to construct an integrated end-to-end QD manufacturing framework that will pave the way for enabling the easy commercialization of QDs.

Research Interests:

As mentioned in the abstract, my research background and interest are majorly focused on the application of process system engineering (PSE) techniques for sustainable energy solutions and my future research interests can be divided into two sections: (a) Multiscale modeling of semiconductor manufacturing processes; and (b) Development of advanced process control techniques to optimize the above process. More importantly, my interest lies in

Specifically, my prior work on QD manufacturing can be extended to other high-demand semiconductors given the supply-chain issues in the US. For example, in silicon processing having high-fidelity lithography and etching model is of great importance. However, the exact computation of each electron and light-ray is computationally expensive, and thus, my interest lies in developing a hybrid model for semiconductor processing that leverages the abundant lithography data and combines it with stochastic electronic KMC (ekMC) simulations. Furthermore, these processes are very sensitive to various inputs and disturbances, and thus, require highly intelligent control frameworks, and current semiconductor control techniques are purely data-driven with custom empirical models for each chipset. Although useful, these techniques are termed ‘black-box’ techniques and do not retain the physics of the process, and thus alternative techniques are necessary. For example, to circumvent this challenge, I am currently working on a Koopman-based control approach, which has a theoretical backing and utilizes decomposition techniques to provide a state-space model that combined with the ML-based techniques to provide a superior controller that retains the physics of the process and provides a hybrid solution. This work will have a significant contribution to the process control community as it combines theoretical fundamentals, with a data-driven ML-based model for the application of high-demand semiconductor manufacturing processes.

Teaching Experience and Interests:

Along with my curiosity and passion for academic research, I also embody an innate zeal toward mentorship and teaching. I view teaching as an essential part of my overall academic activity, and my pursuit of higher education is partly due to these teachers who have inspired me and taught me the necessary skills to achieve my goals and now I wish to do the same - the student must become the teacher!

Specifically, I have embraced the role of a Teaching Assistant (TA) for the ‘CHEN Process Control Course’ and ‘CHEN Semiconductor Microelectronic Processing’. As both courses are relevant to my research, I could deliver the highest value to students in terms of providing real-world examples, and distilling the academic literature to the relevant curriculum for the students. For example, for the ‘Process Control’ course, the Chemical Engineering department implemented experimental control kits in the undergraduate course for the first time, and I was in charge of this section. Here, (a) I could effectively devise a curriculum with 6 experiments and take-home exams, (b) develop the material for this section, (c) teach it to 20+ undergraduate students, and (d) develop well-defined and measurable assessment modules, and it has been a huge success. Furthermore, as my Ph.D. research heavily focuses on modeling and advanced process control, I could correlate the course teaching with the state-of-the-art research, disseminate this information to the students, and provide them with real-world examples to showcase the practical application of theoretical foundations in control theory during my Teaching Assistantship for ‘Process Control CHEN 461’ in Fall 2018 for 80+ undergraduate students.

I believe that my passion and enthusiasm for teaching, my communication skills, and my strong academic background will all make me an excellent teacher. Based on my solid background in chemical engineering, process systems engineering, and process control, I believe that I could comfortably teach various undergraduate courses. My specific research experience inclines me to teach the ‘Reaction Engineering’, ‘Process Control’, and ‘Mass and Energy Balance’. Regarding the graduate curriculum, I plan to develop an ‘Introduction to Machine Learning (ML) Course’ and an elective in ‘Monte Carlo Simulations’. The goal of the ML course would be to familiarize students with various basic and intermediate data-driven techniques like logistic regression and k-nearest neighbors and Support-vector machines (SVM) for data classification, and Neural Networks for regression and black-box modeling. Also, explain the thumb rules, and key insights in the hyper-parameter tuning of these data-driven models, which I think is lacking in many of the ML courses I have experienced. More importantly, I want to demonstrate the above techniques for chemical engineering problems by showcasing results from my research and the literature, to provide the students with a direct application of these abstract techniques. Moreover, dealing with a large amount of data is the norm rather than an exception in today’s world and is an integral part of scientific research. Unfortunately, despite its significance, I generally find a lack of electives in this area within the Chemical Engineering department. Thus, to resolve this knowledge gap, I would jump at the chance to develop such a tailored class with an emphasis on helping students conduct both experimental and computational research, and equip the future students with this important skillset.