TL;DR #
At 100% SOC with 2.0 mm cell spacing, thermal runaway propagates axially through a 21700 module 97 seconds faster than radially, reaching 550.2 °C — a gap that determines whether adjacent cells survive or cascade. For buyers specifying cylindrical cell module layouts, this means inter-cell spacing and SOC operating limits are not packaging preferences but hard safety parameters. Before accepting any 21700 module design, demand documented thermal propagation test data showing axial and radial clearance dimensions, not just cell-level abuse test results.
Overview #
Most buyers evaluate the 21700 format on energy density and form factor alone. That’s understandable — the 21.0 × 70.0 mm cylindrical geometry offers a compelling balance of volumetric efficiency and mechanical simplicity. But when you start looking at module-level thermal behavior, the dimensional choices that feel like packaging details turn out to drive the most consequential safety outcomes in the entire pack design.
The data summarized here comes from controlled thermal runaway propagation experiments conducted at a Chinese engineering university’s mechanical engineering faculty, specifically designed to quantify multi-directional propagation behavior in 21700 module configurations. The test rig — a 1200 mm × 700 mm × 500 mm stainless steel explosion-proof enclosure — used resistance-wire heating (70 W, NiCr wire, 0.5 mm diameter, 7 uniform windings) to trigger a primary cell, with K-type surface thermocouples (±0.1 °C accuracy, rated to 1000 °C) capturing propagation timing and peak temperatures across axial and radial neighbor cells simultaneously. Six experimental conditions varied SOC from 30% to 100% and inter-cell spacing from 2.0 mm to 6.0 mm. The test cell: 21700 format, 5.0 Ah rated capacity, 3.65 V nominal, charged to 4.2 V cutoff at 0.5 C, with discharge cutoff at 2.5 V.
Understanding these results matters for anyone sourcing cylindrical cell formats for pack integration — particularly OEM engineers specifying module geometry or evaluating supplier-submitted designs.
Thermal Runaway Propagation in 21700 Cell Modules: Axial vs. Radial Behavior #
This is where the 21700 format reveals a characteristic that most module designers under-appreciate: the axial thermal conductivity of a cylindrical lithium-ion cell is approximately 20× higher than the radial thermal conductivity. That’s not a marginal difference — it fundamentally changes how heat spreads through a module depending on cell orientation.
In the baseline condition (100% SOC, 2.0 mm spacing), thermal runaway propagated to the axial neighbor cell in just 12 seconds after the primary cell event, with the axial neighbor reaching a peak of 550.2 °C. The radial neighbor took 109 seconds — nearly nine times longer — and peaked at 546.6 °C. At 80% SOC with the same spacing, axial propagation still occurred in 25 seconds versus 145 seconds radially, with peak temperatures of 561.6 °C and 587.6 °C respectively. The axial cell’s thermal runaway trigger temperature (69.6 °C at 100% SOC) was consistently lower than the radial cell’s (80.6 °C), confirming that the axial pathway is simply more thermally aggressive under all tested conditions.
The mechanism isn’t subtle. When the primary cell vents and ignites, high-temperature material ejects directly from the positive terminal along the axial direction — a direct impingement on the neighbor cell’s negative terminal. The radial neighbor, by contrast, absorbs heat primarily through convection and radiation, a slower and more diffuse pathway. Two factors reinforce this: the 20× anisotropy in thermal conductivity, and the physical geometry of the ejection plume.
The four-stage progression observed — heating, thermal runaway trigger (safety valve opening at 128.4 °C at 347 s), full runaway with peak ejection, and burnout — is consistent with what IEC 62619:2022 characterizes as cascading thermal failure in multi-cell assemblies. What this research adds is the directional asymmetry data that the standard doesn’t prescribe design thresholds for.
SOC Effects on Propagation Threshold #
SOC is the single biggest lever buyers and engineers have over propagation risk during storage and transport, short of physical separation.
| SOC Level | Axial Propagation | Axial Trigger Temp (°C) | Radial Propagation | Radial Peak Temp (°C) |
|---|---|---|---|---|
| 100% | Yes — 12 s | 69.6 | Yes — 109 s | 546.6 |
| 80% | Yes — 25 s | 81.5 | Yes — 145 s | 587.6 |
| 50% | Yes — 33 s | 99.5 | No | 105.5 (max, no TR) |
| 30% | No | — | No | 70.7 (max, no TR) |
At 2.0 mm spacing, 50% SOC is the safe radial threshold — radial neighbor temperature peaked at just 105.5 °C, well below any thermal runaway trigger. For axial, 30% SOC is the safe threshold. Below that, even the axial neighbor, despite receiving direct terminal ejection, did not propagate — the primary cell simply didn’t carry enough stored energy to push the neighbor over the edge.
Maximum temperature rise rates follow the same gradient: at 100% SOC, the primary cell hit 40.36 °C/s, axial neighbor 60.7 °C/s, radial neighbor 58.3 °C/s. At 30% SOC, rates dropped to 2.1, 1.3, and 1.7 °C/s respectively — orders of magnitude lower. The implication for transport safety is direct: cells shipped or stored at or below 30% SOC with adequate spacing present dramatically lower cascading risk.
Cell Spacing Thresholds and Safe Distance Requirements for 21700 Module Design #
Inter-cell spacing is the primary design tool for blocking thermal runaway propagation once SOC constraints aren’t viable — for instance, in an operating EV pack at 80–100% charge.
Testing at 100% SOC across three spacing conditions produced clear breakpoints:
| Spacing (mm) | Axial Propagation | Axial Δt (s) | Axial Trigger Temp (°C) | Radial Propagation | Radial Trigger Temp (°C) |
|---|---|---|---|---|---|
| 2.0 | Yes | 12 | 69.6 | Yes | 80.6 |
| 4.0 | Yes | 38 | 89.9 | No | — |
| 6.0 | No | — | — | No | — |
Moving from 2.0 mm to 4.0 mm spacing increased axial propagation delay from 12 s to 39 s and raised the trigger temperature from 69.6 °C to 89.9 °C — the additional thermal path length absorbs enough heat to slow the event significantly. At 4.0 mm, radial propagation stopped entirely. At 6.0 mm, axial propagation also stopped. This gives two usable design thresholds for fully-charged 21700 modules:
- Safe axial distance: ≥ 6.0 mm (no axial propagation at 100% SOC)
- Safe radial distance: ≥ 4.0 mm (no radial propagation at 100% SOC)
Maximum temperature rise rates decreased with spacing: axial cell at 2.0 mm hit 60.7 °C/s, dropping to 53.2 °C/s at 4.0 mm. The radial cell dropped from 58.3 °C/s to non-propagating behavior. At 6.0 mm, peak temperatures for neighbor cells remained at 123.2 °C and 56.3 °C — elevated but sub-threshold.
The physics is straightforward: as spacing increases, the thermal conduction path lengthens, heat dissipates before reaching the neighbor, and the neighbor never reaches its trigger temperature. This is also why the UL 9540A Test Method for Evaluating Thermal Runaway Fire Propagation in Battery Energy Storage Systems focuses so heavily on inter-cell and inter-module spacing as a design variable — because the experimental evidence consistently shows it’s one of the most effective passive mitigations available.
Most procurement teams don’t realize that the safe spacing values for the next-generation 46-series cells (4680, 4695) are expected to be substantially larger than the 4.0–6.0 mm thresholds established for 21700 cells, given the larger stored energy per cell. Research on those formats is still emerging, but anyone designing modules around the 46-series today should not assume the 21700 thresholds are transferable.
Mass Loss Analysis and Physical Evidence #
In supplier qualification, we saw in these experiments that mass loss rate after thermal runaway correlated strongly with SOC — and this is a measurable, verifiable outcome that buyers can use to validate supplier claims about cell abuse tolerance.
At 100% SOC, the primary cell’s mass loss rate exceeded 81.64%, with axial and radial neighbor cells at 75.88% and 66.45% respectively. At 80% SOC, mass loss rates were 81.10%, 74.21%, and 62.10%. At 50% SOC, only the primary and axial cells showed loss (45.90% and 46.60% — the radial cell, having not undergone thermal runaway, showed 0%). At 30% SOC with no propagation, all three cells showed 0% mass loss.
A key detail: axial cells consistently showed approximately 5% higher mass loss than radial cells in all conditions where both propagated. This aligns with the direct ejection impingement effect — the axial neighbor receives more direct energy flux and undergoes a more complete combustion event.
Notably, inter-cell spacing had minimal influence on mass loss rate for cells that did propagate — spacing determines whether propagation occurs at all, but once a neighbor cell reaches thermal runaway, the severity of its destruction is primarily governed by its own SOC.
Practical Guidance for Buyers #
If you’re specifying or evaluating a 21700-based module, the most immediately actionable numbers from this research are the spacing thresholds. For modules operating at or near full charge — any EV application, most stationary storage — axial inter-cell spacing below 6.0 mm means you’re accepting propagation risk. Radial spacing below 4.0 mm carries the same implication. These aren’t worst-case numbers; they’re the minimum distances at which propagation stopped under fully-charged conditions.
Honestly, most buyers over-specify BMS protection parameters while ignoring the physical module geometry. A sophisticated BMS will not stop a thermal runaway cascade once a cell has already vented — the propagation to an axially adjacent cell happens in 12 seconds. The BMS doesn’t act fast enough at that timescale. The geometry has to do the work.
For storage and transport scenarios, the 30% SOC recommendation is well-supported by the data. At 30% SOC with any practical inter-cell spacing, propagation simply doesn’t occur. If a supplier tells you they ship cells or modules at high SOC without documented spacing justification, that’s a procurement risk.
When evaluating module designs, also check which direction the positive terminal faces relative to neighbors. Axial alignment means direct ejection exposure — that’s the 12-second propagation pathway. Radial alignment buys you time, but not infinite time: at 2.0 mm and high SOC, radial propagation still occurs within 109 seconds.
At CompactBESS, our sourcing team works with verified Chinese manufacturers of cylindrical cell modules and complete pack assemblies, connecting OEM buyers and integration engineers with suppliers who can document thermal management geometry and inter-cell spacing to these standards. If you’re evaluating 21700 module designs or qualifying suppliers for pack integration, we can help you ask the right questions before samples arrive.
Need help identifying qualified suppliers for 21700 cylindrical cell modules with documented thermal spacing compliance? Talk to our sourcing team →
Supplier Qualification Questions #
- What is the axial inter-cell spacing in your 21700 module design, and can you provide thermal runaway propagation test data confirming no axial propagation occurs at that spacing under 100% SOC conditions — specifically the propagation delay time (Δt1) and trigger temperature for the neighboring cell?
- At what SOC level does your module design achieve zero thermal runaway propagation in both axial and radial directions at your stated inter-cell spacing, and how was this threshold validated — using direct heating at 70 W or equivalent, with K-type thermocouple surface measurement at ±0.1 °C resolution?
- What is the maximum temperature rise rate (°C/s) measured at neighboring cells during thermal runaway testing at your design’s operating SOC range, and how does this compare between axial and radial neighbors?
- For 21700 cells in your module, what is the documented thermal runaway trigger temperature for adjacent cells at 2.0 mm spacing versus your production spacing, given the experimental baseline of 69.6 °C (axial) and 80.6 °C (radial) at 100% SOC?
- Can you provide post-thermal-runaway mass loss rate data for your 21700 cells at 100% SOC, and what is the measured percentage — given that experimental data establishes a >80% mass loss rate threshold for primary cells at full charge, which directly indicates whether cell-level containment is sufficient?
Sourcing Checklist #
- ☐ Module design specifies axial inter-cell spacing ≥ 6.0 mm for applications operating at 100% SOC, verified by dimensional drawing or physical measurement
- ☐ Module design specifies radial inter-cell spacing ≥ 4.0 mm for applications operating at 100% SOC, verified by dimensional drawing or physical measurement
- ☐ Supplier provides thermal runaway propagation test report showing propagation delay time (Δt) for both axial and radial directions under relevant SOC conditions, with thermocouple data logged at ≥1 s intervals
- ☐ Storage and transport documentation specifies SOC ≤ 30% for 21700 cells, consistent with zero-propagation threshold established at 2.0 mm spacing
- ☐ Thermal runaway test setup includes simultaneous axial and radial neighbor monitoring, not single-direction testing only — confirming multi-dimensional safety evaluation per IEC 62619:2022 requirements
- ☐ Cell specifications confirmed as 21700 format (21.0 × 70.0 mm, 5.0 Ah class), with rated capacity, nominal voltage (3.65 V), and charge/discharge cutoff voltages (4.2 V / 2.5 V) matching procurement specification
- ☐ Supplier can document that neighboring cells show ≤ 140 °C peak temperature (non-propagating threshold) under design operating conditions, consistent with test data at 30% SOC
Key Specifications Table #
| Parameter | Recommended Value | Verification Method |
|---|---|---|
| Axial inter-cell safe spacing (100% SOC) | ≥ 6.0 mm | Dimensional inspection + thermal runaway propagation test at 100% SOC; confirm no axial trigger in neighbor cell |
| Radial inter-cell safe spacing (100% SOC) | ≥ 4.0 mm | Dimensional inspection + thermal runaway propagation test at 100% SOC; confirm no radial trigger in neighbor cell |
| Maximum safe SOC for storage/transport (2.0 mm spacing) | ≤ 30% | Battery charger/discharge tester verification at 0.5 C, confirm by OCV measurement post-static |
| Axial neighbor thermal runaway trigger temperature (2.0 mm, 100% SOC) | > 69.6 °C (baseline threshold) | K-type surface thermocouple at positive-terminal end, 1 s logging interval, during 70 W resistive heating test |
| Radial neighbor thermal runaway trigger temperature (2.0 mm, 100% SOC) | > 80.6 °C (baseline threshold) | Same thermocouple method; confirm trigger temperature exceeds baseline for design spacing |
| Cell format dimensions | 21.0 mm diameter × 70.0 mm length | Caliper measurement, batch sampling per incoming inspection protocol |
| Primary cell mass loss rate at 100% SOC | > 80% (indicates full venting event) | Weigh cell before and after thermal runaway test; ratio confirms completeness of event |
Can’t find a supplier meeting these specs? Submit your requirements and we’ll match you within 48 hours.
References #
Data source: Multidirectional Thermal Runaway Propagation Characteristics of Cylindrical 21700 Lithium-Ion Battery Modules: Effects of State of Charge and Inter-Cell Spacing, E.-X. Pan et al., Journal of the Electrochemical Society, 2025
Frequently Asked Questions #
Why does thermal runaway propagate faster in the axial direction than radially in a 21700 module?
Two mechanisms work together. First, when a cylindrical cell undergoes thermal runaway, it ejects high-temperature material directly from the positive terminal — meaning any cell aligned axially receives a direct impingement of hot gas and debris on its negative terminal. Second, the axial thermal conductivity of a cylindrical lithium-ion cell is approximately 20 times higher than the radial conductivity, so heat travels much faster along the cell axis than across its diameter. At 100% SOC and 2.0 mm spacing, this produces a 97-second difference in propagation timing between the two directions.
What is the minimum safe inter-cell spacing for a fully-charged 21700 module?
For cells operating at 100% SOC, the experimental data establishes 6.0 mm as the safe axial distance and 4.0 mm as the safe radial distance. Below these thresholds, thermal runaway propagation was observed under test conditions. These values apply to the 21700 format specifically — do not assume they transfer to larger-format cells like the 4680 series.
Can SOC management alone eliminate thermal runaway propagation risk without increasing spacing?
At 2.0 mm spacing, reducing SOC to 30% stopped all propagation in both axial and radial directions. At 50% SOC, radial propagation was eliminated but axial propagation still occurred (33 s delay, peak temperature 446.2 °C for the axial neighbor). So SOC management is effective but the safe threshold differs by direction — 30% for axial safety, 50% for radial safety, at tight spacing. For storage and transport, the 30% limit covers both.
Does inter-cell spacing affect how severely a cell is damaged once thermal runaway does occur?
No — and this is an important nuance. Spacing determines whether propagation happens, but once a neighboring cell reaches its thermal runaway trigger temperature, the severity of its destruction is governed primarily by its own SOC. Mass loss rates for cells that did propagate were similar across spacing conditions. The 100% SOC primary cell consistently showed >80% mass loss regardless of neighbor spacing.
How do these thermal spacing requirements relate to compliance standards for battery modules?
The IEC 61960-3 standard for secondary lithium cells and batteries covers cell-level performance but does not specify module-level inter-cell spacing thresholds. The thermal propagation spacing data from this research is most directly applicable to module design validation under UL 9540A and to safety assessments under IEC 62619:2022, which requires that battery systems be designed to prevent propagation from a single-cell event to adjacent cells. The specific spacing thresholds — 6.0 mm axial, 4.0 mm radial — provide quantitative design targets to underpin that compliance claim.
Published by compactbess.com Technical Team | Request a sourcing quote
