John Bonini (Fed)

I work on the development and application of first principles and data-driven methods for the modeling and prediction of properties in material and chemical systems. Broadly speaking the aim is to connect the microscopic structure of a system to the properties which make it useful and interesting. Materials are often useful or "functional", when they respond strongly or in a particular manner to external changes in things like electromagnetic fields, temperature, and strain. I have worked on methods for modeling dielectric, piezoelectric, ferroelectric, and hall effect (thermal and electronic) properties. These material properties are essential in the functionality of various sensors, motors, memory, and other devices.

Identifying and tuning systems to realize these functional properties can enable the development of next generation devices. I work to facilitate this by developing first principles based methods to model systems where magnetic, electronic, vibrational, and other processes couple to one another, particularly when the relevant energy and time scale for such processes are not well separated. This has included magnon-phonon coupling in strong SOC systems and strong light-matter coupling in optical cavities yielding "polaritonic" excitations. I'm also interested in multi-scale approaches where models of larger length scale properties are constructed using parameters obtained from smaller more computationally accessible simulations. This enables both simulation of a wider variety of systems represented in a more realistic manner and investigation into how the larger scale structure (which can be more readily controlled in synthesis) can be used to tune systems to optimize functional properties. I'm especially interested in approaches to treat systems where electronic properties play an important role at larger length scales. Previous work has included modeling of dielectric properties of perovskite oxide superlattices and the large length scale electronic structure which can occur in twisted "Moiré" systems.

The methods used in my work are based on first principles quantum simulations. Predominantly the starting point is density functional theory (DFT) calculations which may be used to compute system properties directly (e.g. via density functional perturbation theory) or combined with downfolding methods such as Wannierization to construct effective models. Lately I have been working to combine these established physics informed methods with tools that have enabled recent practical advances in machine learning. Specifically, incorporating automatic differentiation with existing methods seems a promising route to facilitate both extending the physics we can model as well as efficiently quantifying the sensitivity of our results to simulation parameters.

NRC Postdoctoral Fellowship 2024
Flatiron Research Fellowship 2020
Fundamental Physics of Ferroelectrics Workshop Presentation Award 2019