What "thermal runaway-free" means

The paper is Zhang and colleagues, "Thermal runaway-free ampere-hour-level Na-ion battery via polymerizable non-flammable electrolyte," published in Nature Energy on 6 April 2026. The last author is Yong-Sheng Hu of the Institute of Physics, and CarNewsChina reported that the work is linked to HiNa, a sodium-ion battery company spun out of the same institute. (The paper's own competing-interests statement says the authors declare none.)

The idea is clever. The electrolyte, which the team calls PNE, is a liquid in normal use. When the cell gets hot, it polymerizes into a cross-linked solid that sits between the electrodes, keeping them from reacting with each other and cutting down the gases that feed a runaway. The abstract's claim is direct: the design led to "no thermal runaway in ampere-hour-level cells," and nail-penetration tests were passed without smoke, fire or explosion.

The coverage went further. CarNewsChina wrote that the results indicate full interruption of thermal runaway pathways. New Atlas described "batteries that simply cannot catch fire" and said the design eliminates the possibility of a fire. Those are different claims, and the difference is the whole point of this report.

What the heating test actually was

The main heating result is an accelerating rate calorimeter (ARC) test, the same instrument behind Report 228. An ARC heats a cell in small steps and waits to see whether the cell starts making its own heat. If it does, the instrument follows it, and you learn the temperature where self-heating starts and how fast it runs away.

Supplementary Figure 14 shows that test on the PNE cell: a 3.5 Ah sodium-ion cylinder in the 26700 size, charged to 3.95 V before testing. Its temperature climbs steadily for roughly 3,500 minutes, more than two days, up to about 300 °C, and then falls as the run ends. There is no sharp spike. The figure is labeled "No thermal runaway," and the paper's comparison table describes the cell as having passed the 300 °C ARC test.

That is a genuine result. It also has a ceiling built in. A test that runs to 300 °C can tell you the cell did not run away below 300 °C. It cannot tell you what happens above that, in a fire that is already burning next to the cell. "Survived a test that stopped at 300 °C" and "cannot catch fire" are not the same sentence. (That reading of the curve is mine; the figure itself does not say why the run ended where it did.)

The best detail is the control the headlines skipped

The paper tested two comparison electrolytes alongside PNE, and Supplementary Figure 17 plots how fast each cell heated itself in the ARC:

E3, a conventional carbonate electrolyte: self-heating between roughly 170 and 260 °C, peaking around 130 °C per minute.
E5, a phosphate-solvent electrolyte: self-heating from about 140 °C, with spikes past 2,000 °C per minute, by far the most violent of the three.
PNE: a flat line at zero across the whole range.

(Those values are my readings off the plot, not numbers printed in the paper.)

Now look at Supplementary Table 2. E5 and PNE use the same two salts at the same concentrations. The only difference is the solvent: trimethyl phosphate (TMP) in E5, triethyl phosphate (TEP) in PNE. Both belong to the phosphate family of solvents that people reach for when they want an electrolyte that will not burn. One version produced the calmest cell in the study and the other produced the most violent. In a separate heating test of the electrodes alone, the supplementary discussion reports that the anode reacted with E5 so violently, and made gas so fast, that it vented the sealed test crucible.

That is the paper's real lesson, and the abstract says it outright: the link between how flame-retardant an electrolyte is and how safe the battery is remains unclear. "Non-flammable electrolyte" is a property of a liquid in a dish. Thermal runaway is a property of a whole cell, electrodes and all. The headline "cannot catch fire" runs exactly the shortcut this paper warns against.

The 211 Wh/kg number came from a different cell

Every story I read paired the safety result with an energy density of 211 Wh/kg, as if one cell did both. Supplementary Table 4 gives that calculation, and it is for a smaller cell: a 1.85 Ah, 29.3-gram cylindrical pouch cell with a different cathode (listed as SC-NFM424) and a 3.34 V mid-point voltage. The ARC result is for the 3.5 Ah 26700 cell. The paper's comparison table lists 1.45 Ah and 3.5 Ah cylindrical cells as the safety-tested formats.

The supplementary file does include a nail-penetration video of the 211 Wh/kg cell charged to 4.3 V, so that cell was abused too. What I did not find in the supplementary material is an ARC heating run on it. Two different cells, two different sets of tests. That is ordinary in a materials paper. It just should not be compressed into one battery that is both the most energy-dense and the most fireproof.

What this means for a home or grid battery

A home battery or grid container is hundreds or thousands of cells, wired, cooled and packed together. The question that matters at that scale is whether one failing cell can set off its neighbors, which is why system-level fire testing exists (see Report 101 on what UL 9540A does and does not certify). A single-cell ARC run and a nail test are where that story starts, not where it ends. Sodium-ion's march toward grid storage is covered in Report 011, and the case for a chemistry that cannot burn at all, a water-based zinc battery, is in Report 030.

If a product launches on the strength of this electrolyte, the questions to ask are simple: was the cell you are buying the cell that was tested, at what charge level, and has the full module or system been through a propagation test? Those are my questions, not the paper's, but they are exactly the gaps between what was measured and what was headlined.

The honest limits on this

The main text of the paper is behind a paywall, and I did not read it. Everything above comes from the abstract, the article page, and the open supplementary file (its 22 figures, 4 tables and discussion), which I read in full. I did not watch the ten supplementary videos, so I rely on their titles and on the abstract for the nail-test outcomes. I could not confirm from what I read how many cells were tested per condition, or whether 3.95 V is the full-charge voltage for the 3.5 Ah cell. The main text may answer some of this; if it changes the picture, I will correct this report in place. None of this is a knock on the chemistry. Stopping a runaway in an ampere-hour cell is hard, and the TEP-versus-TMP comparison is a useful result in its own right.

This is close to my own work: I help design the AI battery-cycling systems for a veteran-owned (HUBZone) energy-storage integrator, run by fellow veterans. I do not own that company and earn nothing from this link. Full policy here. Nothing in this report involves them or any product they sell.