Same lithium in all three. The only difference is the dissolved salt — the part everyone treats as passive plumbing. Hit Run plating and watch the metal build. One of these is quietly courting disaster.
A team led by Zhiyuan Zeng at City University of Hong Kong built a sealed cell thin enough to sit inside an electron microscope — a 35 nm imaging window over a sliver of liquid — and filmed lithium forming on an electrode, second by second. Earlier cells used windows near 100 nm over a ~1,000 nm liquid layer, which blurred lithium's first moments.
Same lithium, three different salts. The negatively charged partner — the anion — turned out to steer how the metal grew:
The honest scope: none of this surprised battery scientists — fluoride‑rich films had long been tied to longer cell life. What no one had done was watch the protection form in real time. It's a measurement upgrade: from photographing the aftermath to filming the event.
This is lithium‑metal battery research — the higher‑energy next generation, not the lithium‑ion cell in your phone. The tests were coin cells at lab scale. So this isn't a "better battery" you can buy; it's a clearer picture of a known mechanism.
The payoff is two methodological things. A screening method: you can now judge a candidate salt by watching lithium grow under it, instead of building full cells and waiting. A design rule made concrete: grow a hard inner LiF layer to block spikes and let lithium creep flat, under a soft outer layer that flexes without tearing — and you tune the salt's anion to get exactly that.
The deeper lesson, and the reason this lab exists: what you can see decides what you can know. For years the two‑layer film was inferred from leftover surfaces and models. A thinner window turned a strong hunch into something you can watch. That's the whole move.
Primary source: Mei, Zeng et al., “Operando identification of anion effect on lithium nucleation and growth via in‑situ transmission electron microscopy,” Nature Communications (Zhiyuan Zeng’s group, City University of Hong Kong; published 13 June 2026, doi:10.1038/s41467-026-74340-1). Popular write‑up via Earth.com.
Real, from the paper: the 35 nm window vs ~100 nm prior; the three growth modes (dendritic / moss / sideways‑fusing sheet); the two‑layer SEI with a ~20 nm LiF shell under a few‑nm soft skin; the coin‑cell head‑to‑head (LiTFSI > 500 h and ~92% efficiency vs the fluorine‑free salt shorting in under half the time at under 40%).
Stylized here: every shape on the canvas is a hand‑drawn schematic, not a solver. Timing, branch counts, and the tilt are chosen for legibility. The fluorine‑free salt's exact identity isn't named in the popular write‑up, so it's shown only as “fluorine‑free.” No numbers are computed from the animation — the lab illustrates the mechanism and quotes the paper's measurements.
The pedagogy: names stripped first (Lester's Method). You commit to a pattern, then receive the labels, then meet the exception that breaks the rule you built yourself.