The failure signature came first
Before there was a root cause, there was a pattern. NHTSA’s Office of Defects Investigation opened Preliminary Evaluation PE 20-016 on October 9, 2020, covering 2017–2020 model year Chevrolet Bolt EVs. The complaints described fires starting under the seating area in parked, unattended vehicles. All of them alleged the fire happened during or after electrical charging. ODI’s review found the thermal events occurred during and/or at the end of the battery module charging cycle (ODI Resume, PE 20-016).
That is the diagnostic tell, and it is worth sitting with before the recall numbers. A fire that clusters at end-of-charge is not a random cell failure. It is a defect that becomes dangerous when the cell is pushed to its highest state of charge, where the anode is most expanded and the mechanical clearances inside the jelly roll are at their tightest. The failure mode was pointing at a physical defect inside the cell long before anyone named it.
What GM and LG found inside the cell
The ODI resume states the root cause directly: General Motors and LG Energy Solution, working with LG Electronics, determined the thermal events resulted from two simultaneous manufacturing defects occurring in the same battery cell. The two defects, located within the same cell module, caused an internal short within the battery pack, which produced the thermal events. The manufacturing process was changed to prevent both conditions.
The ODI document describes the defects generically. The specific names came from LG. A spokeswoman for LG told Consumer Reports that the cause was a torn anode tab and a folded separator within the battery modules, and that the defects had been addressed in manufacturing. Consumer Reports attributes that identification to GM’s explanation and the LG spokeswoman’s confirmation.
Confidence note: the primary ODI filing confirms “two simultaneous manufacturing defects in the same cell” but does not name them. The torn-tab/folded-separator identification rests on the Consumer Reports account quoting LG. The full defect information reports (21V-560 and 21V-650) would settle the naming from the primary filing; those PDFs were not retrievable for this piece.
Why the pair matters more than either defect alone
A torn anode tab is a current-collector integrity problem. A folded separator is a dielectric barrier problem. Either one alone is a manufacturing escape worth screening for. Together, in the same cell, they are a different category of hazard: a marginal internal clearance where a conductive path can form across what should be an insulator, at the point in the charge cycle where mechanical stress is highest.
This is the part that gets lost in recall summaries. The population-level defect rate can look vanishingly small — both defects have to land in the same cell — while the per-cell consequence remains catastrophic. That asymmetry is why sampling-based confidence was never going to be an adequate remedy, and why the corrective action ended up being module replacement rather than inspection.
The recall timeline, in order
The sequence matters because each step revised the scope based on new field data.
- October 9, 2020 — ODI opens PE 20-016 on 2017–2020 Bolt EVs after complaints of fires under the seating area in parked, unattended vehicles, all during or after charging.
- April 29, 2021 — GM files Recall 20V-701. The remedy: diagnostic procedures to identify and replace potentially defective cell-module assemblies, plus onboard diagnostic software to detect and warn owners of changes in battery module performance over time. Until that interim remedy was available, GM instructed owners to enable Hilltop Reserve or Target Charge Level to cap charging at 90 percent.
- July 24, 2021 — GM files Recall 21V-560, covering all 2017–2018 Bolt and certain 2019 Bolt vehicles with design level N2.1 cells produced at the LG Ochang, Korea plant. Remedy: replace defective cell modules.
- August 21, 2021 — GM expands the scope. Additional analysis led GM and LGES to conclude the manufacturing defects were not isolated to one battery manufacturing plant. Recall 21V-650 covers all 2019–2022 Bolt EV and 2022 Bolt EUV vehicles.
All of the above is from the ODI resume. The expansion is the load-bearing fact for anyone doing supplier qualification: a defect characterized as plant-specific turned out to be process-specific. The factory address was the wrong unit of analysis.
The charge-state aggravation, and what it implies for screening
ODI’s language is precise: the defect condition “appears to be aggravated by routinely charging the battery to a full or nearly full state of charge after it has been substantially depleted.” That is a stress-profile statement, not a defect statement. It says the latent defect is present regardless, but the duty cycle that drives the cell to full charge from a deep depletion is what converts it into a thermal event.
This is why the interim guidance was a 90 percent state-of-charge cap plus a request not to deplete below roughly 70 miles of remaining range. Those two instructions are the same physics from opposite ends: keep the cell away from the top of its charge window, and avoid the deep-swing-then-full-recharge pattern that stresses it most.
For qualification engineers, the implication is uncomfortable. A cell screening program that only tests at mid-state-of-charge electrical parameters will not see this class of defect. The defect lives at the mechanical and dielectric level, and it manifests at end-of-charge. A screen that never drives cells to full charge under a realistic duty cycle is testing the wrong operating point.
Software was a detector, not a remedy
The 20V-701 diagnostic layer deserves a fair reading. It was genuinely useful: ODI states that the information obtained from the 20V-701 software update was part of what allowed GM and LG to determine the root cause. The monitoring generated the field data that identified the defect pair. That is a real contribution.
But a monitoring layer cannot remove a physical defect. It can flag an abnormality in battery module performance over time and warn the owner. It cannot un-tear an anode tab or unfold a separator. The actual corrective action was hardware: module replacement. Consumer Reports reported GM’s plan as replacing all modules in 2017–2019 Bolt EVs and replacing defective modules in 2020–2022 Bolt EV/EUVs, with the pack case, electronics, and wiring retained because they were not defective.
The design lesson is not that software monitoring is worthless. It is that monitoring and remediation are different functions and should never be conflated in a corrective action plan. A detector buys time and generates evidence. It does not close the defect.
What the standards do and do not reach
Pack-level abuse testing standards exist for good reason, and SAE J2464 is the reference most often cited for electric vehicle battery abuse testing. But the retrieved page for SAE J2464 returned only the site tagline, not the standard’s scope text, so this piece will not characterize its specific test content. What can be said without overreach: abuse testing at the pack level is designed around defined mechanical, thermal, and electrical insults. A latent manufacturing defect pair that produces an internal short only at end-of-charge under a specific duty cycle is a different failure pathway, and it is not obvious that any pack-level abuse test is constructed to catch it.
That is the standards gap worth naming. The Bolt incident was not a design-abuse failure. It was a manufacturing-escape failure that survived into the field and only manifested under a particular charge profile. The control that would have caught it lives upstream, at the cell manufacturing line and the incoming inspection bench, not at the pack abuse test chamber.
Bench-verifiable rules that follow
These are recommendations, not findings from the recall documents. They are the rules I would write into a cell qualification plan after reading the ODI resume and the LG identification.
- Screen tab welds and separator placement mechanically, not just electrically. Incoming lot inspection should include magnification or X-ray/CT checks on anode tab integrity and separator fold/placement. A torn tab and a folded separator are physical features; they are visible to the right instrument. Electrical test at mid-SOC will not reliably surface either.
- Add an end-of-charge stress step to qualification. Because the failure signature clusters at end-of-charge after deep depletion, qualification should include a duty cycle that drives cells to full charge from substantial depletion while logging self-discharge and temperature. Mid-SOC-only screening tests the wrong operating point.
- Treat a plant-specific defect finding as process-specific until proven otherwise. The 21V-650 expansion happened because the defect was not isolated to Ochang. Supplier qualification should audit the manufacturing step and its process controls, not the factory address. A corrective action scoped to one plant is a hypothesis, not a closure.
- Keep monitoring and remediation in separate columns. Onboard diagnostics are a detector. They generate evidence and buy time. They do not remove a latent physical defect. Any corrective action plan that lists software as the remedy for a hardware defect should be sent back.
- Log formation and end-of-charge data at the cell level and retain it. The 20V-701 software data was what let GM and LG find the root cause. Cell-level formation and end-of-charge logs are the manufacturing-side equivalent, and they are cheap relative to a fleet recall.
FAQ
What exactly was wrong with the Bolt cells?
Per the ODI resume, two simultaneous manufacturing defects in the same cell caused an internal short. Per LG, quoted by Consumer Reports, the two defects were a torn anode tab and a folded separator. The primary filing confirms the pair but does not name them; the naming rests on the LG statement.
Why did the fires happen during or after charging?
ODI found thermal events occurred during and/or at the end of the charging cycle, and that the defect condition is aggravated by routinely charging to full or near-full after substantial depletion. End-of-charge is where the cell is mechanically and electrochemically most stressed, which is consistent with a latent internal defect becoming a short.
Did the software fix solve the problem?
No. The 20V-701 diagnostic software was an interim measure that generated the field data used to find the root cause. The corrective action was hardware module replacement under 21V-560 and 21V-650. Software detected; hardware remedied.
Why did the recall expand from one plant to all Bolts?
GM and LGES determined the manufacturing defects were not isolated to one battery manufacturing plant, which is why 21V-650 expanded to all 2019–2022 Bolt EV and 2022 Bolt EUV vehicles. The defect was a process problem, not a site problem.
What should a qualification engineer take from this?
Screen the physical defect at incoming inspection, stress cells at end-of-charge during qualification, audit the process rather than the plant, and never let a monitoring layer stand in for removing a latent hardware defect.
