How Ask Battery Experts works

How an answer is made

1. Find the pages. Your question is matched against the text of every page of each researcher's papers, and the dozen most relevant pages are picked.
2. Answer only from those pages. A language model (Claude, by Anthropic) writes the answer in the first person, as a voice for describing the published work. It isn't allowed to add opinions, anecdotes or unpublished results.
3. Cite every claim. Each statement carries a numbered reference to the paper and page it came from, with the quoted passage, so you can check it.
4. Say when the papers don't cover it. If the pages don't answer your question, the persona says so instead of guessing.

What the AI can get wrong

• The AI can misread a paper, combine findings that don't belong together, or miss later work that changed the picture.
• Answers only reflect the papers in the collection, which may not include every paper, or the newest ones.
• "I" and "we" describe the published work. They aren't the researchers' own words or current views.

For anything that matters, open the cited page and read it yourself.

Who answers

• Virtual Anna (AI)

  Built from 30 full-text papers by Anna Stefanopoulou (University of Michigan). Covers expansion/force sensing, electrode state of health, formation and SEI, parallel-cell imbalance, venting and thermal-runaway detection, second-life screening.

  Summary and papersEmail the real Anna

• Virtual Greg (AI)

  Built from 71 full-text papers by Greg Offer (Imperial College London). Covers physics-based and degradation models, cell cooling and thermal gradients, degradation mechanisms, pack design, lithium-sulfur cells, and electrochemical-thermal testing.

  Summary and papersEmail the real Greg

• Virtual Venkat (AI)

  Built from 54 full-text papers by Venkat Viswanathan (University of Michigan). Covers lithium-metal and solid-state electrodeposition and dendrites, electrolyte design with machine learning and autonomous experiments, electrochemical kinetics beyond Butler-Volmer, SEI interfaces, and batteries for eVTOL aircraft.

  Summary and papersEmail the real Venkat

• Virtual Birger (AI)

  Built from 51 full-text papers by Birger Horstmann (DLR / Helmholtz Institute Ulm). Covers SEI growth and aging models, chemo-mechanics of silicon anodes, transport in concentrated and ionic-liquid electrolytes, lithium stripping and plating, zinc batteries, and Bayesian parameterization of physics-based models.

  Summary and papersEmail the real Birger

• Virtual Ziyou (AI)

  Built from 17 full-text papers by Ziyou Song (University of Michigan). Covers state-of-health and lifetime prediction from field data, physics-informed machine learning, integrated thermal and energy management for electric vehicles, fast-charging control, and second-life batteries.

  Summary and papersEmail the real Ziyou

• Virtual David (AI)

  Built from 71 full-text papers by David Howey (University of Oxford). Covers Bayesian parameter estimation and model selection, electrochemical and nonlinear impedance spectroscopy, state and temperature estimation, degradation knees and lifetime prediction with machine learning, fault detection in parallel packs, and grid-scale storage.

  Summary and papersEmail the real David

• Virtual Valentin (AI)

  Built from 25 full-text papers by Valentin Sulzer (Ionworks). Covers physics-based model hierarchies and asymptotic reduction, PyBaMM and pack simulation, degradation and knee modelling, formation and lifetime prediction, and verifiable battery modelling practice.

  Summary and papersEmail the real Valentin

Ionworks guide (AI)

Not a researcher: an AI guide to the Ionworks battery guide (fundamentals, modelling, parameterization) and blog, built the same way and citing the section it draws on. It answers only when you pick it, never as a best fit, so the researchers' answers stay about their published work. Researchers listed here don't use or endorse Ionworks by appearing next to it.

Open the guide

PyBaMM (AI)

Not a researcher: an AI guide to the PyBaMM documentation (user guide, example notebooks and API reference) and to 66 open-access papers whose authors used PyBaMM, citing the docs section or paper page it draws on. It answers when you pick it or name PyBaMM in a question, never as a best fit for other questions, so the researchers' answers stay about their published work. The papers' authors and the researchers listed here don't endorse anything by appearing next to it.

Open the PyBaMM docs

The personas are not the real researchers, who haven't reviewed or endorsed their answers. To ask the real researchers, use the email links above or next to each persona in the chat.

Missing someone? Suggest a researcher. Wrote a paper a persona can only cite by title, or know where to find a free copy? Help us add it.

Question limits

Every question is answered by a large language model, which costs money to run, so asking is limited for visitors who aren't signed in: 5 questions a minute, 30 questions an hour and 100 questions a day. One question counts once, however many personas answer it. A shared daily limit also applies across all visitors, so on a busy day questions can pause for everyone.

While you wait, the glossary is always open.

Your questions and privacy

• Stored anonymously. We keep each question, the answers it got and the country it came from, but not your IP address or browser details. Questions from the same visitor can be grouped within one day, and can't be linked to you or across days.
• Seen by Ionworks and the researchers. The Ionworks team reads the questions to improve the personas, and each researcher may see the questions their persona answered, with its answers.
• Leave out personal or confidential details. Your question's text is stored as you wrote it, so anything you put in it is kept.
• Deleted after 365 days. Older questions and their answers are removed automatically.
• Answered by Claude. Your question and the conversation so far are sent to Anthropic's API to write the answers.

On your own site

Anyone can put a chat with any persona on their website: a button with the researcher's photo in the bottom-right corner opens a panel that answers only from their papers, with the same citations, limits and privacy as here. It takes one line of code.

• Get the code. Copy it from the persona's page: Virtual Anna, Virtual Greg, Virtual Venkat, Virtual Birger, Virtual Ziyou, Virtual David, Virtual Valentin.
• Choose the suggested questions. Up to three, written into the code, shown before anything else.

Who built it

Ask Battery Experts is built by Ionworks, the battery data and simulation layer.