Welcome too the Physics-Informed Energy Materials Lab (PIEML))

We build computational models of how energy materials work and how they fail, from the motion of
individual ions up to the response of a whole device. The governing equations for these systems
are known. What is missing is a way to solve them at the scales where experiments are actually
performed, which is why the field falls back on empirical fits that carry no mechanism.

Our approach is to write the physics as differentiable programs. The same code then predicts
forwards, recovers mechanism backwards from sparse experimental data, and returns a design,
because gradients pass through converged solutions. Machine learning enters as a way to solve and
invert governing equations rather than to replace them, which keeps the models thermodynamically
consistent, keeps their parameters physically meaningful, and removes the need for the large
labelled datasets that materials problems rarely have.

We work on batteries and solid electrolytes, solid oxide electrolysis cells and fuel cells,
molten salts for advanced nuclear reactors, high-entropy alloys, and nanoporous materials for
carbon capture and hydrogen storage.
 

Key Research Themes

We write the governing equations of electrochemistry and materials science as differentiable programs, so that one code predicts forwards, recovers mechanism backwards from sparse measurements,
and designs. And we measure how much the data actually determine, so every number we report comes with a statement of whether it is supported.

Differentiability supplies the capability. Identifiability supplies the brake. That pairing is what makes five areas one research program.

Physics-informed and differentiable solvers

The equations are stiff, coupled, and on moving boundaries. Standard neural solvers stall on them by orders of magnitude, and we build the ones that do not.

Electrochemical interfaces, electrodes and corrosion

From an image of a real electrode to a predicted spectrum, with no equivalent circuit anywhere.

Statistical mechanics of fluids and interfaces

One free energy, differentiated automatically, from gas separation to battery interfaces to freezing.

Atoms: forces, structure and inverse design

Getting the atoms right, because everything downstream inherits the error.

Inverse problems and identifiability

What a measurement can determine, in corrosion, degradation and anomalous transport.

Join us

We take students and postdocs from chemistry, physics, materials science, chemical engineering and applied mathematics. You do not need all of it: a background in electrochemistry and a willingness to learn scientific computing works as well as the reverse. If you write code and want to work on problems where the answer has to satisfy a conservation law, get in touch at cgtetsas@uwaterloo.ca with a CV and a paragraph on what you would like to work on.