Page 20 - IJAMD-1-2

International Journal of AI for
Materials and Design
Sustainable electronics using AI/ML

technologies with unprecedented performance and chemical sustainability discovery acceleration involves
functionality. addressing two key factors: securing access to relevant,
The primary difficulty in new material discovery lies in large datasets in chemical research and promoting open
111,112
analyzing the materials data with numerous dimensions. data for fair and equitable AI development. Efforts
To address this challenge, ML techniques are utilized, are underway to create an open-access framework
adjusted, and enhanced. There are several ML methods and infrastructure for organizing and disseminating
that can be used based on the task of the experiments. organic reaction data through a centralized repository,
109
These methods are classification, regression, clustering, a crucial step for advancing AI in the future. Thus,
dense neural network, sequential learning, dimension the transformative capacity of AI extends to the realm
reduction, variational autoencoders, convolutional of chemical engineering and chemistry, encompassing
neural network, and BO. The AI algorithms analyze the advancements in the exploration of innovative materials,
validation of results to enhance experimental strategies, optimization of processes, and enhancement of quality and
thereby enabling more precise predictions for subsequent safety standards.
experiments. Unlike in automated laboratories, where 8. Summary and perspectives
researchers preset experimental procedures, this closed-
loop approach is adopted. Consequently, autonomous This paper contains a concise review of the recent
laboratories could refine traditional trial-and-error advancements in transient technology, specifically in the
methods, accelerating the discovery of chemicals and context of transient electronics, with a focus on analyzing
materials. Moreover, they might mitigate the trade-off the materials, designs, and performances of these devices.
between the cost of human-driven testing, which may A major part of this review focuses on transient materials,
have reached its peak in terms of efficiency gains, and the encompassing metals, polymers, and semiconductor
expected time to discovery. Besides, AI-driven quantum materials. Further research should target enhancing the
mechanics and computing resources have the capability efficiency of the developed transitory devices so that they
to revolutionize predictive toxicology, an underlying can be on comparable levels with conventional devices.
aspect of sustainability analysis. 110 While there is a trade-off between the performance of a
device and its transitory capabilities, it is crucial to achieve a
Through advanced algorithms and ML techniques, suitable balance. The development of AI and robotics-based
AI shifts through vast databases of existing materials, molecular design tools for autonomous high-throughput
scientific literature, and experimental data to experimentation is accelerating. These technologies hasten
uncover hidden patterns, correlations, and potential the process of generating and optimizing commercially
breakthroughs. AI-driven approaches leverage vast viable materials and equip chemists with tools for
amounts of data and advanced algorithms to streamline predicting molecular properties. This versatile and data-
113
and enhance the process. Material discovery using AI driven approach is required to address the time-consuming
is a multistage task, mainly involving four key stages: and expensive procedure brought on by the complications
Characterization, property prediction, synthesis, of the experiments. For large-scale bulk applications, it will
and theory paradigm discovery. The initial stage be essential to be able to precisely and logically forecast
involving gathering information about a material how variations in molecular-level characteristics translate
lays the foundation for subsequent discovery phases. to bulk properties. Because of the higher degrees of
114
The characterization stage encompasses imaging and variability and possible combinations of repeating units,
instrument-related contexts and creates vast high- extending data-driven technologies from optimization
dimensional data, which surpass human processing of small molecules to polymers unveils challenges for the
capabilities. Thus, AI can assist by automating and effective identification of desirable options. As a result, the
enhancing the characterization process, reducing effective use of AI will pave the way for the discovery of
manual labor, enhancing data quality, and uncovering novel molecular designs and the repurposing of existing
valuable insights from complex datasets. The imperative molecules for novel applications. Building a cohesive
115
to revolutionize the approach to transient electronics and complete data collection of attributes is crucial for the
stems from the critical need to mitigate environmental advancement of AI and automation for the discovery of
impact, paralleling the requirement for the development novel materials. This includes defining and standardizing
of eco-friendly chemicals to align with sustainability systems representing polymers used in ML models. More
objectives and minimize the gap between technological advancements in ML techniques are required to realize
advancement and ecological responsibility. the potential benefits of data-driven polymer chemistry,
AI heavily depends on extensive, high-quality databases including discovery and chemical synthesis simplification.
crucial for software learning. Overcoming roadblocks in After structuring the data, it is necessary to compile

Volume 1 Issue 2 (2024) 14 doi: 10.36922/ijamd.3173