Is an aged battery more dangerous?

The question comes up every time a home battery passes its warranty, and every time someone considers a second-life pack built from retired EV cells. If the battery now holds 80% of what it did new, is it more likely to fail badly, or less?

The authors of the second study point out that most battery safety research has been done on fresh cells, possibly on the assumption that aged ones will behave about the same or less severely because they store less energy. So two groups, at Sandia National Laboratories and at ZSW in Ulm, Germany, set out to check. They aged commercial 18650 cells in two different ways and then abused them: overcharge, overdischarge, nail penetration, crush, and slow heating in an accelerating rate calorimeter (ARC). Their results came out in the Journal of The Electrochemical Society in 2025, as Part I and Part II of one series.

Two ways to lose capacity

A lithium-ion cell loses capacity mainly by losing the lithium it shuttles back and forth. The two papers separate two ways that happens.

SEI growth. A thin film called the solid electrolyte interphase grows on the graphite electrode over the cell's life, and it locks up lithium as it grows. Part II aged its cells under conditions that made this the dominant mechanism.

Lithium plating. Instead of slotting into the graphite, lithium deposits on its surface as metal. Part I produced this by cycling cells at 0 °C. The authors note that those conditions were within the specification the cell manufacturer gave.

Both routes end at the same headline number. A cell can reach 80% state of health either way. That is the point that matters for everything below.

Aged by plating: it heated itself from about 35 °C

Part I (Gerosa and colleagues at ZSW) tested 29 commercial 3 Ah NCA cells, new ones against ones cycled at 0 °C until they carried plated lithium, which the team confirmed by electrochemical tests and by opening cells up afterward.

In the ARC, which heats a cell in small steps and watches for it to start generating its own heat, both plated cells began self-heating at about 35 °C. The new cells began at about 80 °C. The plated cells also heated themselves faster, and across the overcharge, overdischarge and overcurrent tests they reached higher hazard levels on the standard EUCAR scale. The authors trace this to plated lithium metal reacting with the electrolyte, and they stress it was not caused by lithium dendrites shorting the cell.

Thirty-five degrees is a warm garage, not an abuse test. That is why this finding matters more than any single peak temperature: the reaction that starts a runaway was already running at a temperature a battery can sit at in normal life.

Aged by normal wear: about the same as new

Part II (Preger and colleagues, Sandia and ZSW) repeated the exercise on three chemistries, NCA, NMC and LFP, aged so that SEI growth dominated, with no plating. Here the authors' overall reading of the heating tests was that, on kinetics, the safety profiles of the aged cells are unchanged or slightly safer. Their summary across every test: the aged and new cells showed a similar magnitude of thermal runaway, but how easily runaway was triggered sometimes shifted.

For the LFP cells, which are the chemistry most home batteries use, the numbers were:

Self-heating onset (ARC): 105 °C new, 105 °C at 80% state of health. No change.
Venting temperature: 157.7 °C new, 156.4 °C aged. No real change.
Peak temperature (ARC): 374 °C new, 450 °C aged. Higher.
Overcharge: the aged cell took 24.3 minutes to trip its internal cutoff against 16.7 minutes for the new one, but reached 92.3 °C against 48.1 °C, vented, and leaked electrolyte.
Overdischarge: essentially the same (peak 34.8 °C against 33.9 °C).

So even with the "good" kind of aging, the LFP result is not a clean "safer." The temperature at which trouble starts did not move, but the aged cell ran hotter once it went, and it handled overcharge worse on the hazard scale. The NCA cells aged this way needed less nail penetration to trigger runaway (2.8 mm against 3.2 mm, with far less force), another case where an aged cell was easier to tip over even though it failed less violently afterward. The authors' own conclusion is that the differences between chemistries were larger than the differences between new and aged cells of any one chemistry.

Why the capacity number can't answer this

Put the two papers side by side. Part II's 45 °C cycling (no plating) and Part I's 0 °C cycling (heavy plating) were run on the same NCA cell type and reached a similar state of health. One set began self-heating at about 98 °C, slightly later than new cells. The other began at about 35 °C. Same cell model, similar remaining capacity, a difference of more than 60 degrees in where the trouble starts.

That is the honest answer to the headline question: an aged battery is not reliably safer or more dangerous. It depends on the aging mechanism, and the usual health readout, remaining capacity, does not tell you which mechanism you have. The Sandia paper's introduction summarizes earlier work the same way: plating has consistently lowered onset temperatures by up to 50 °C, while cells that lost lithium only to SEI growth looked like fresh cells or slightly better.

What this means if you own one

The Part I authors draw a practical line from their own results: avoid the conditions that cause plating, which they name as critical combinations of low temperature, high charging rate and high state of charge, and check for plating before reusing cells in second-life systems. In plain terms, the risky history is a battery that was charged hard while cold. A battery that simply got old in a mild climate is, on this evidence, the less worrying case.

For a home system, that puts the weight on the battery management system and on where the unit lives. A battery that can be charged hard and full in a freezing garage would be exposed to the combination these authors warn about (that is my application of their advice, not something they tested). How a battery fades over time is the subject of Report 205, and heat's side of the story is in Report 114. If you are weighing reused EV cells, Report 089 covers that market.

The honest limits on this

These are small numbers of small cells. Part II tested one LFP cell per condition with no replicates, because the team did not have enough cells, and the LFP cells were 1.1 Ah cylinders, not the large prismatic cells inside a home battery. The plating result comes from NCA cells; neither paper tested LFP cells aged with plating, so I am not claiming the 35 °C figure carries over to LFP. Every cell was charged to 100% before testing, a deliberate worst case. Part I's own authors say their 0 °C cycling produced large amounts of plated lithium that may not represent real use, and that the amount of plating needed to change safety results is not yet known. The authors also flag cell format, capacity and how cells are assembled into modules as factors still to be studied. The "warm garage" comparison and the side-by-side of the two NCA results are my reading of the two papers together.

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.