(home)

Research

Here, I describe my current research interests. Some of these relate to my main academic posting in "Digital Chemistry" at UCD, while others are far afield. For information about my research group, check out the CoReACTER!

(Directed) Hypergraph Theory

Hypergraphs are generalizations of graphs. Whereas in graphs, edges encode binary relationships, connecting always two vertices, in hypergraphs, the analogous hyperedges can encode relationships involving arbitrary numbers of vertices. Though hypergraphs can always be represented as biparite graphs, there are a number of applications where hypergraphs are the more natural object of study. These include chemical reaction networks (CRNs, see below), where chemical reactions can be represented as directed hyperedges from the reactants (tail) to the products (head).

In spite of their utility, hypergraphs have not received nearly as much attention in the fields of mathematics and computer science as graphs. Directed hypergraphs, in particular, have been neglected in the literature. I've been slowly trying to help remedy this situation. In addition to developing open source codes for hypergraph analysis, I've been working on formalizing undirected and directed hypergraph theory and doing some mathematical and statistical analysis on CRN-like directed hypergraphs.

Three reactions are shown, r1: M and N go to P and Q; r2: M and P go to T; and r3: 2Q go to U. Each reaction and each species is given a different color. This small reaction network of six species and three reactions is represented as a directed hypergraph, where vertices (chemical species) are represented by circles filled with the color corresponding to one of M, N, P, Q, T, or U and hyperedges (reactions) are represented by curved arrows, some of which have multiple tails or heads, given a color corresponding to one of r1, r2, or r3. one vertex (T) is labeled 'species (vertex)', and similarly, one hyperedge (r3) is labeled 'reaction (hyperedge)'.

Machine Learning on Chemical Reaction Networks

In recent years, interest in data science and machine learning (ML) in the chemical sciences has exploded. It's now common to use graph-based machine learning to predict the properties of molecules, materials, and even chemical reactions. However, there's been comparatively few developments applying ML to predict the properties of chemical networks.

I'm currently working to develop tools for chemical reaction network machine learning, or CRN-ML. We won't know until we try, but I believe that CRN-ML could help us to more efficiently predict species and reaction properties while also unlocking totally new capabilities that are only possible with network-level analysis.

A directed hypergraph is represented by a series of black circles with white interiors (vertices) connected by curves arrows, some of which have multiple tails or heads (hyperedges). The text 'CRN' is next to this hypergraph. Below, a 'reaction representation' is obtained by passing some reaction (the example given is A goes to B) into a neural network (represented by a series of interconnected circles), the output of which gives a hyperedge feature vector (represented by a stack of red, yellow, and orange squares). Similarly, above the CRN hypergraph, a 'species representation' is obtained by passing molecules or materials into a neural network, yielding a vertext feature vector represented by blue and blue-green squares. The CRN, with these vertex and hyperedge features, is passed (represented by an arrow) into a neural network labeled 'hypergraph neural network', and then this hypergraph neural network is connected (again by an arrow) to 'link prediction' (where a smaller, gray version of the same CRN hypergraph is shown, but with an added hyperedge. The added hyperedge is pink, represented by dashed rather than solid curves, and has a question mark above it, suggesting that it might or might not exist) and 'regression' (where a plot of predicted versus reference Gibbs free energy barriers is shown, with a parity line and a series of pink circles scattered around that line).

Mechanistic Analysis in Electrochemistry

"Rational design" has become a buzzword of late. Chemical scientists want to be able to design optimal, purpose-built molecules and materials based on consistent design rules, rather than discovering such compounds serendipitously or through brute-force trial and error. Relatedly, it is a long-term goal for researchers to be able to rationally design synthesis procedures to create these desired, often novel compounds.

While rational synthesis planning and even complete retrosynthesis is possible in many areas of conventional organic chemistry, rational design remains a far-off goal for electrochemistry, where reactivity is often extremely complex, often involving reactions in multiple phases and depending sensitively on the electrode, supporting electrolyte, and electrochemical conditions (i.e., potential and current) used.

I'm interested in advancing our understanding of electrochemistry and electrosynthesis using a variety of tools from computational chemistry and data science. My prior research in this area has included conventional quantum chemistry and atomistic molecular dynamics simulations as well as mesoscale and continuum-scale analyses informed by elementary reaction mechanisms. I continue to be interested in fundamental and multi-scale modeling of electrochemical reactivity, with cross-talk between anode and cathode being an area of particular focus. CRN analyses and machine learning (either as a method to predict reactive and transport properties or as a method to accelerate kinetic and dynamic simulations) supplement my modeling efforts.

A cartoon depiction of an electrochemical cell, specifically a Li-ion battery. The negative electrode (a gray background with lighter gray circles, suggesting graphite) has some material (purple and green circles) deposited on it, and a molecules (an organic carbonate consisting of red, black, and white circles) is on top of this deposited material. Some other molecules are nearby in the electrolyte, which is represented by a blue background, and these molecules are suggested to be reacting at the surface by a curved arrow. Similarly, on the right, there is a positive electrode (a green background covered with purple, red, and green circles, meant to suggest a transition-metal oxide), and we see some electrolyte molecules again reacting, forming some other small molecules (carbon dioxide, represented by two red circles connected to a black circle in a line; hydrogen fluoride, represented by one green and one white circle; a difluorophosphate, represented by two green and two red circles - one connected to a white circle - all connected to a central orange circle; and lithium, represented by a purple circle). The processes on the two sides of the cell are shown to be linked by two large arrows forming a loop. Around the circles are the words 'chemical kinetics' (next to the Eyring equation), 'mass transport' (next to an expression for flux), and 'electrochemical kinetics' (next to a rate coefficient equation from Marcus theory).

Scientific and Chemical Ethics

Research and applications in the chemical sciences have immense moral consequences. While chemistry has the power to save lives, as exemplified by the development of stable vaccines, pharmaceuticals, and fertilizers to improve agricultural yield (among other chemical developments), it has also been a force for tremendous destruction, whether intentional (e.g., chemical weapons; militarized chemicals such as Agent Orange; the use of Zyklon B in the Nazi genocides against those who were Jewish, Roma, queer, and/or disabled) or unintentional (e.g., tetraethyl lead as an additive in gasoline causing mass brain damage; greenhouse gas emissions from combustion and other chemical processes causing climate change; plastics poisoning the environment and human and non-human organisms). Chemists must carefully weigh the value and the dangers of the work that they pursue and the substances they design, create, and use, yet the ethics of chemistry are neither widely discussed in the scholarly literature nor widely taught to chemistry students at the undergraduate or graduate levels. This is a receipe for disaster in the gravest sense.

My interest in scientific and chemical ethics began amid the recent hype around so-called "artificial intelligence" ("AI") and foundation models. I was (and remain) concerned that scientists were increasingly turning to tools that I believe(d) to be dangerous and in fact harmful to scientific research and education. This led to my study on the scientific and chemical ethics of foundation models. As part of this study, I conducted a grounded theory analysis of chemical codes of ethics/codes of conduct, synthesizing from them a notion of chemistry's stakeholders and the ethical obligations of chemists to those stakeholders.

I continue to be interested in scientific and chemical ethics, including applied and professional ethics related to emerging technologies in and around chemistry. I am also actively engaged in pedagogical projects to teach chemistry students to think critically about ethical issues in research and educational settings.

Cartoon scales in yellow are surrounded by various words written at all different angles: 'AI', pedgagogy, epistemic (in)justice, climate, publications, equity, intellectual property, corporate interests, belonging, consent, wellbeing, pollution, open science.

Science, Madness, and the Self

There is a growing understanding that there is a "mental health" crisis in science. The social and professional environment scientists work in is frequently toxic and even traumatizing; even when scientists avoid the worst outcomes, scientific work culture often values productivity over wellbeing, and the hypersane, hyperrational philosophy espoused by many scientists make even conversations about emotional and cognitive states challenging. My interest in madness (a non-pathologizing and more expansive term than the sanist "mental illness") began from this point, leading to questions such as: What is the relationship between science and madness? Are the two totally incompatible, or do they/can they coexist? How could science change (or, how could something replace science) to embrace and be inclusive of Mad ways of knowing?

At present, I'm trying to combine science and technology studies with neurodiversity studies and Mad studies to understand the madness-science relationship. I'm also engaged in (auto)ethnography related to experiences of and relating to madness. My work in this area is in the very earliest stages, madness a curiosity; it might lead to nothing, or it might be incredibly fruitful and generative.

A black-and-white picture of a femme face. They are crossing their eyes, and a hand - possibly their own - is squeezing their cheeks. The right side of their face is covered with small stickers of stars and crescent moons. Image by Jorge Rojas taken under the Unsplash license.