SATPC0034775 Tab 04 4 SOW..pdf

Prototype implementation of Bayesian Optimization (BO) for Electrohydrodynamic Ink Jet

Printing

Lars Kotthoff; Hantang Qin

Project Description

Long-duration exploration missions require a paradigm shift in the design and manufacturing of space architectures. The ability to perform In-Space

Manufacturing (ISM) provides a solution towards sustainable, flexible missions (both in-transit and on-surface) through on-demand fabrication, repair, and recycling capabilities for critical systems, habitats, and mission logistics and maintenance. For example, electronic components used in space missions are exposed to harsh environments; therefore, advancing

ISM capabilities for electronic device fabrication and repair is essential.

My team is proposing a high-risk and transformative research project to enable ISM of semiconductors and electronic devices – a capability that is critical to making long-duration space flights by humans a reality. Current methods for semiconductor fabrication are impractical in space as they require large facilities and complex multi-stage process steps such as masking, lithography, etching, deposition, and thin film growth. The proposed work aims to develop Electrohydrodynamic (EHD) Printing technology as a promising alternative process providing a non-contact (defect reduction), direct printing (mask-less) method, and etching-free process for semiconductor manufacture. To date, most 3D printing techniques cannot be operated in a gravity-free environment because gravity is one of the dominating factors for successful 3D printing that relies on mechanisms such as liquid or powder flow. EHD Printing utilizes electrical force to drive liquid ink deposition and achieve 3D printing in gravity-free environments in the micro and nano scale. EHD printing offers a flexible solution to creating conductive structures for electronics, such as sensors, displays, transistors, optical devices, and more (Fig.1). Traditionally the electronic components of these technologies have been fabricated using lithography based methods. However, these processes offer poor customizability, little prototyping ability, and require costly equipment within a clean room environment. These are limitations EHD printing overcomes, providing a fast, versatile manufacturing method for fabricating electronics in-space.

Fig. 1. Examples of EHD printed 2D/3D polymer/metal in µm

Prelim Work of the Research Team

The collaborative research team has been teaming with NASA to work towards On-Demand Manufacturing of Electronics (ODME).

Our goal is to realize thin-film fabrication for the International

Space Station (ISS) and other On-orbital Services, Assembly and

Manufacturing (OSAM) needs. Upon success, we can provide manufacturing capabilities for ISM of all different types of semiconductors and electronic devices. This could be an essential tool to make sure America can maintain long-term missions and human presence in low earth orbital with a strong commercial market considering chips/sensors are most likely to be consumables for in-space uses. Previously, the research team has finished two three-day zero-g parabolic flight tests in Fort Lauderdale, FL. To our knowledge, this was the first time that EHD printing was validated in a microgravity environment. The research team has successfully demonstrated that

EHD printing is a unique gravity-free 3D printing technique (see Fig.2). Conductive silver patterns and perovskite ink patterns have been successfully printed.

Current Challenges and Research Goals

EHD printing applies an electrical force through the ink via a conductive nozzle to create a Taylor

Cone. The Taylor Cone allows the ink to overcome the surface tension at the nozzle and be drawn into droplets with a smaller diameter than the nozzle itself. It allows small droplets to pinch off from the meniscus creating conductive patterns as small as 500nm to 1μm. To successfully achieve this phenomenon, EHD printing requires a precise set of process parameters: flow rate, applied voltage, nozzle/substrate stand-off distance, platform speed, and specific ink characteristics

(conductivity, viscosity, surface tension, etc.). Currently, EHD printing requires process parameters to be manually set prior to printing. Additionally, quality control has been completed after printing to detect any defects. The goal of this proposed study is to enable remote sensing and control in EHD printing for ISM based on AI-enabled schemes.

Bayesian Optimization for Robust EHD Printing

The state-of-the-art paradigm for optimizing complex processes like the EHD printing process is

Bayesian Optimization. In a nutshell, Bayesian Optimization is an iterative, machine-learning-driven process that can leverage small amounts of data to quickly optimize a process whose inner workings aren’t known, even if evaluations of this process are noisy. We will optimize across the parameter space described above to identify optimal settings for a given environment. Crucially, the result will not be a single static setting that will only work in a specific case, but a robust process that can be deployed to perform the optimization again in a variety of environments.

Our previous work (Hud Wahab et al., Ceramics International, https://doi.org/10.1016/j.ceramint.2023.12.225) applied the same methodology to optimize laser-induced graphene production. The figure shows that Bayesian Optimization is able to improve the outcome (the conductivity of the produced graphene, measured through the resistance) even with a small number of experiments – substantial improvements over the initial results start to appear after < 20 experiments. Being able to work with small data is especially important in this context, as the number of experiments we can run is limited. We propose to apply this tried-and-tested

Fig. 2. Parabolic flight test by my research team – EHD printing validation for ISM https://doi.org/10.1016/j.ceramint.2023.12.225 methodology to the EHD printing process and implement it in a way that it can be used in space to optimize in-place for the specific environment.

Roles and responsibilities

In this project, Prof. Kotthoff at the University of Wyoming will take the lead, supported by Prof.

Qin at the University of Wisconsin Madison, who will lead the experimentations through a separate contract with NASA. In a lab environment, multiple initial printing tests on a new material will be conducted. The initial data sets will be used for the Bayesian Optimization training. With the suggested optimized system inputs, the second round or multiple rounds of optimization experimentations will be conducted to validate the effectiveness of optimization. Prof. Kotthoff’s team will focus on the algorithm development specifically targeting the EHD printing process. The team will publish at least one research article on this effort.

Project Schedule (anticipated period of performance 5/15/24-9/15/24):

- month 2: definition of parameter space to optimize, what instruments to integrate with

- month 4: prototype that optimizes the agreed-upon parameters of the EHD printing process and evaluation of prototype to judge efficacy of the optimization and sample efficiency

Key deliverable: Code and initial empirical results for applying BO for EHD printing of flexible electronics, plus a final report on the performed activities at the end of the project.