Research
What happens inside a battery, and how I watch it
A plain-language tour first. Open the technical sections if you want the details.
The big problem
The world needs a lot more energy storage, and it needs to be cheap and safe. Rechargeable zinc batteries with water-based electrolytes are a promising option: water doesn’t burn, and zinc is abundant. But they wear out, and much of the reason is hidden in a thin, busy layer where the zinc electrode meets the liquid.
One of the biggest problems is dendrites: as zinc plates back onto the electrode during charging, it can grow in sharp, needle-like spikes instead of a smooth layer. Dendrites waste zinc and can eventually pierce through to the other electrode and short the battery. My thesis, in the Hsu Group, looks at reversible epitaxial zinc electrodeposition: getting zinc to grow in an orderly way that follows the crystal pattern of the surface beneath it, cycle after cycle, so dendrites don’t form.
The questions I’m answering
- What controls whether zinc plates back as an orderly, epitaxial layer or as dendrites?
- What forms at the zinc–electrolyte interface as a battery charges and discharges, and in what order?
- How can what we learn change the way battery cells are designed?
TODO(Sidney): confirm or replace these questions.
The approach
I watch the battery while it runs (operando) using Raman spectroscopy, which reads molecular fingerprints from scattered laser light. To see the first few nanometers of the surface, I use SHINERS: gold nanoparticles in ultra-thin glass shells that amplify the signal right where the chemistry happens. Because no off-the-shelf cell does everything these measurements need, I design and 3D-print my own.
What I built
- Custom 3D-printed spectroelectrochemical cells, designed and iterated in-house.
- Python tooling for spectral analysis that turns long series of spectra into something people and models can reason about.
- A machine-learning workflow, with Yuxin Chen’s group, that compares simple and complex models and is heading toward choosing experiments automatically.
Results to date
Results from this work aren’t public yet. Papers and talks will appear here as they come out.
Publications: thesis papers in preparation. Earlier work: an undergraduate poster at the ACS Spring Meeting (2022) on computational screening of water-oxidation photocatalysts (see CV).
Threads
Go deeper
- 01Operando Raman spectroscopyWatching a zinc battery while it worksShining a laser into a running zinc-ion battery to see which molecules form and disappear at the electrode surface as it charges and discharges.
- 02SHINERSSeeing the first few nanometersUsing gold nanoparticles wrapped in an ultra-thin glass shell to amplify the signal from the very surface of an electrode, where the important chemistry happens.
- 03Custom 3D-printed spectroelectrochemical cellsCells built for the laserDesigning and 3D-printing my own battery test cells so a laser can look inside while they run, and iterating on the design in days instead of weeks.
- 04Machine learning collaboration with Yuxin Chen's groupTeaching a model to pick my next experimentBenchmarking simple and complex machine-learning models on my experimental data, then using active learning so the model helps choose which experiment to run next.● In progress. Comparing ridge regression, random forests and neural networks; active learning is next.