(2df) Autonomous Labs to Accelerate Discovery and Understanding of Organic Mixed Conducting Materials

Organic mixed ionic-electronic conductors (OMIECs) are amongst the most promising materials for the next generation of electrochemical devices. Their potential applications in areas such as healthcare and energy means that these materials will be topics of intense interest for decades to come. These complex materials present new fundamental questions that autonomous labs are uniquely positioned to answer more efficiently (Figure 1). My research group will develop autonomous labs for studying OMIEC materials. We will leverage the strengths of automation and machine learning (ML) to drive important scientific discoveries. The two main focuses of my group will be improving our understanding of structure-property relationships in mixed conduction organic materials and developing tools for automated and data-driven experiments. For these goals, I will recruit group members with a broad range of interests, including materials, automation and ML. A core tenet of the group will be open science. We will work toward this goal by using preprint servers, public repositories for code and data, and sharing the tools we develop with the scientific community.

Proposal 1: Inverse Design of High Entropy Organic Mixed Ionic-Electronic Conductors

Most organic thin films are made from only a few components. Automated experiments will allow us to easily explore the landscape of highly complex OMIEC thin film formulations. ML can navigate these multi-dimensional landscapes much more efficiently than conventional approaches. My group will use our automated lab to achieve very fine control over the structure and properties of OMIECs using complex multi-solvent processing and multi-component blends. Advanced ML techniques will guide inverse design of OMIEC materials and processing to achieve state-of-the-art ionic and electronic conductivity.

Proposal 2: An Automated Lab for Organic Thin-Film Materials

Automated experiments could produce more reliable structure-property relationships and free up researchers for higher level scientific inquiry (Figure 2). Preparing thin film materials requires careful control of experimental and environmental variables, which can be reliably performed by robots. We will build a robotic system for automated fabrication and characterization of OMIEC thin films. Using automated and manual experiments, we will construct “maps” of structure-property relationships in OMIECs, and generate large datasets thereof. The automated experimentation systems developed by my group, along with proof-of-concept demonstrations, will advance the progress of automated research and lead to its more widespread adoption.

Proposal 3: An AI Agent for Scientific Understanding of Organic Films and Devices

Researchers use experiential or intuitive understanding to select e.g. the appropriate solvent, concentration and deposition parameters to transform a single polymer or blend of polymers into a film or device with useful properties. Artificial intelligence (AI) systems require significantly more training data than most researchers. The key advantage of human researchers is that they can extrapolate to unknown systems by generating scientific theories and understanding, which can be used to make qualitative predictions without the need for complex calculations. My group will develop models than can be used to extract quantitative scientific understanding and theories on the relationship between molecular structure, processing and thin-film/device properties. We will leverage the capabilities of our automated lab to generate large quantities of data, as well as curating and releasing datasets from the literature. Initially, we will develop ML models than can explain what they have learned (e.g. feature importance, model uncertainty, etc). Finally, we will use ML models to extract quantifiable molecule-processing-structure-property (MPSP) relations in the form of useful equations through means such as extracting complex differential equations and symbolic regression (equations that relate molecular structure and processing to properties).

Teaching Interests

Training the next generation of scientists is one of the most important jobs a faculty member can undertake. In my own work, I will try to achieve this by teaching both undergraduate and graduate courses, as well as through mentoring members of my research group. One of my favorite aspects of teaching is helping students to develop not only their understanding of the course material, but also critical thinking and planning skills. I believe that developing students and scientists who can effectively approach and solve any problem is more important than simply ensuring that they have mastery of the subject material.

In my roles as a teacher and supervisor, I use questions to gauge a student’s understanding of a topic. Students who thoroughly understand a subject should be able to explain an idea clearly without resorting to overly technical terms or jargon. I strive to create classroom and group environments where students understand that they are expected to be highly prepared and participate regularly, but also will feel comfortable enough to express when they are struggling to understand something. In such instances, I find it extremely valuable to guide them through the thought process – either in class or during office hours – that will lead them to a solution without directly indicating how the problem should be solved. I developed this pedagogical style as a result of my own experiences as a student. Classes that involved a significant amount of memorization or rote learning were often more difficult and less fulfilling than those where students were expected to demonstrate their understanding of the concepts. However, I also recognize that I am only at the early stages of my growth as a teacher, and so am very keen to continue my development in this area using evidence-based teaching methods.

Demonstrating how course material is relevant outside of the classroom is a powerful way to engage students more deeply. I will regularly bring in examples from the literature or from my own group to illustrate the importance of a particular topic. This, too, stems from my own previous experiences. In my undergraduate quantum mechanics course, many of the students, including myself, were very excited to have a homework assignment based on a 2000 Physics Review Letters paper on the bond length and binding energy of helium dimers. The fact that I remember it to this day is a testament to the impact of tying course material to real-world examples. This can also be easily extended to outreach activities by using fun examples such as the complex crystallization behavior of chocolate to demonstrate the relevance of chemistry or materials science.

As a faculty member, I look forward to teaching courses in areas including design, synthesis, processing and characterization of polymers and materials, as well as introductory courses in chemical engineering. I see this pedagogical work as having significance in the longer term as a means for preparing the next generation of scientists. As industry and academia continue to embrace the power of data science and automation, it will be important for the department to provide students with education in these emerging areas. There are genuinely exciting opportunities for both classroom- and lab-based courses that focus on the tools of 21^st century research: coding, automation and data science. As a member of your faculty, I would like to help lead this expansion. For example, I would like to teach a lecture course that introduces students to the Python programming language, and teaches them to use it for data analysis and visualization, computation and machine learning for research. Additionally, I believe that a laboratory course designed to teach students how to set up, control and troubleshoot automated experimental systems would be very exciting, and could turn into a very successful course. Simple syringe pumps could be used for this course. On the other hand, a robot like the Opentrons OT-2 is a highly capable and inexpensive system that would allow students to tackle more complex experiments.