Electric vehicles have experienced explosive growth in recent years, driven by global pressure to reduce carbon emissions and to secure sustainable energy supply. Among the available electrochemical energy-storage devices, lithium-ion batteries offer the best combination of high specific energy, long cycle life and low self-discharge rate. They have consequently become the dominant building block of modern traction battery pack designs. However, the compactness and energy density that make lithium-ion cells attractive also create severe hazards under abuse conditions. A local mechanical, electrical or thermal abuse event can release stored energy within a few seconds, heating adjacent cells and causing a cascading thermal-runaway propagation. In a crowded traction battery pack, this process very often leads to a violent fire or even an explosion that consumes the entire vehicle. To guarantee the safety of electric vehicles, therefore, it is essential to study not only the ignition phase of a single cell but also the ways in which thermal runaway spreads in the surrounding cell population.
The work presented here concentrates on a commercial 18650 cylindrical cell with a LiNi0.8Co0.15Al0.05O2 cathode, commonly used in automotive-grade traction battery pack modules. The whole investigation followed a combined experimental and numerical route. First, a dedicated heating-rod experiment was built inside a constant-temperature chamber to trigger thermal runaway by thermal abuse and to record the cell voltage, surface temperature and visible phenomenology. Then ceramic-fibre silica aerogel composite boards were inserted between the heating source and the cell to evaluate their capability to mitigate thermal propagation. Because the manufacturer supplied aerogel boards in discrete thicknesses, a simulation model was validated first and then used to identify an optimum single-layer thickness. The selected single-layer insulation was subsequently installed in a small battery module, and the numerical results confirmed that it was capable of protecting the surrounding cells. Finally, to improve the barrier further without adding considerable thickness, a three-layer structure composed of a 4 mm aerogel sheet with thin metal skins was proposed and tested. Both the metal-layer thickness and metal material were varied to find the configuration with the strongest thermal-blocking performance.
1. Heat-Transfer Background for Thermal-Runaway Propagation
Thermal runaway propagation in a traction battery pack is fundamentally a heat-transfer problem. The three governing modes are conduction, convection and radiation. Fourier’s law is usually written as
$$ \Phi = -\lambda A \frac{\partial T}{\partial x} \qquad\qquad (1) $$
where \(\lambda\) is the thermal conductivity, \(A\) the area normal to the heat-flow direction and \(\partial T/\partial x\) the temperature gradient. Convective heat transfer between a surface and the surrounding fluid follows Newton’s law of cooling,
$$ \Phi = h A \Delta T \qquad\qquad (2) $$
with \(h\) representing the surface heat-transfer coefficient. Thermal radiation is particularly important during battery thermal runaway because the radiator temperature reaches several hundred degrees Celsius; the Stefan-Boltzmann law for a real surface reads
$$ \Phi = \epsilon A \sigma T^{4} \qquad\qquad (3) $$
where \(\epsilon\) is the surface emissivity and \(\sigma\) the Stefan-Boltzmann constant. During the present experiments, heat was supplied by an externally heated stainless-steel rod. The cell, insulation layers and heater exchanged heat through conduction in the solid regions, natural convection driven by buoyancy in the surrounding air, and radiation across the gaps. The contribution of each mode was quantified indirectly in the simulations.
From a calorimetric viewpoint, the thermal-runaway of a lithium-ion cell can be approximated by four distributed reactions: SEI decomposition, anode-electrolyte reaction, cathode decomposition and electrolyte decomposition. The total heat-generation rate is
$$ \dot{Q}_{total} = \dot{Q}_{SEI} + \dot{Q}_{anode} + \dot{Q}_{cathode} + \dot{Q}_{ele} \qquad\qquad (4) $$
and each sub-process is governed by an Arrhenius-type kinetic law, for example:
$$ \frac{d\chi_{SEI}}{dt} = -A_{SEI} \,\chi_{SEI} \exp\left( -\frac{E_{a,SEI}}{RT} \right) \qquad\qquad (5) $$
Such kinetic descriptions are required when a simulation must predict the onset of thermal runaway. In the present study, however, the cell surface temperature remained substantially lower than the decomposition range in most protection tests. Consequently, the numerical model ignored the internal chemical heat sources and treated the cell as a passive solid body. This simplification was justified by the relatively low peak temperatures reached when the aerogel insulation successfully blocked the heat flux.
2. Experimental Platform and Instrumentation
A repeatable experimental platform was designed specifically for the safety evaluation of a traction battery pack. The core of the platform was a square constant-temperature chamber with internal dimensions of 0.5 m × 0.5 m × 0.5 m. The chamber was maintained at 27 °C before every test. Inside the chamber, linear rails, sliders and locking fixtures held the heating rod, the test cell and the insulation board at prescribed relative positions. A photograph of the structural arrangement and its principal components is presented in the embedded visual.
The test object was an 18650 cylindrical lithium-ion cell with a nominal capacity of 3000 mAh and a nominal voltage of 3.6 V. The charge cut-off voltage was 4.2 V and the discharge cut-off voltage was 2.0 V. Its main properties are listed in Table 1. Each cell was charged to 100% state of charge before the experiment using a Land battery tester. Cells were first discharged at 1 C to 2 V, rested for one hour, charged at constant current with 0.5 C to 4.2 V, and then charged at constant voltage until the current fell below 150 mA.
Table 1 — Main specifications of the tested 18650 cell
| Property | Value |
|---|---|
| Cathode material | LiNi0.8Co0.15Al0.05O2 |
| Anode material | Graphite |
| Diameter × height (mm) | 18 × 65 |
| Nominal capacity (mAh) | 3000 |
| Nominal voltage (V) | 3.6 |
| Mass (g) | 46.6 ± 1.5 |
| Operating temperature (°C) | 0–60 |
Temperature was measured by K-type thin-patch thermocouples with a measurement range from −40 °C to 700 °C. The thermocouples were bonded to the cell surface with high-temperature tape at three vertical locations, namely near the top, at the middle and near the bottom of the cell. Their average value was taken as the representative surface temperature facing the heater; since the cell shell is thin and its lateral conductivity is much higher than the internal effective radial conductivity, this average is a good measure of the hottest cell surface. The voltage of the cell was continuously recorded by a Keysight 34411A digital multimeter through nickel tabs welded to the electrodes. A thermal camera also monitored the entire process through the sapphire window of the chamber, recording visible flames, jetting and smoke release.
The external thermal-abuse source was an electric cartridge heater with a heated length of 90 mm and an outer diameter of 16 mm. The nominal heating power was 350 W and a PID power controller supplied DC electricity. In all tests the distance between the heater and the cell was set to zero, i.e., the heater, the insulation board and the cell were pressed against each other whenever an insulation structure was present. The cell was located at the centre of the chamber, 100 mm above the bottom plane. The centre height of the 90 mm heated zone was aligned with the middle of the 65 mm cell height. The complete measurement chain is summarized in Table 2.
Table 2 — Overview of test conditions
| Parameter | Value |
|---|---|
| Heater power (W) | 350 |
| Heater diameter × heated length (mm) | 16 × 90 |
| Horizontal spacing between cell and heater | 0 (direct contact in protective tests) |
| Cell bottom height above chamber floor (mm) | 100 |
| Heating duration (s) | 1200 or until thermal runaway |
| Chamber ambient temperature (°C) | 27 |
3. Thermal-Runaway Characteristics of the Unprotected Cell
In the first series of experiments, a fully charged 18650 NCA cell was heated without any passive barrier until thermal runaway was triggered. A representative temperature and voltage evolution curve is shown in the open literature of this type; the recorded data in the present study showed the same distinctive four-stage process. The process can be summarized as follows:
Stage I — Gradual temperature increase. The cell surface temperature rose continuously under the external heating. No obvious voltage change was observed, although a small transient voltage increase followed by a minor decrease was visible. SEI decomposition started near the heated side, but the associated heat release was not large enough to change the global heating trend.
Stage II — Onset of internal short circuit (ISC). When the cell surface temperature reached about 190.74 °C, the separator began to melt locally. This caused a sudden voltage drop. The voltage then fluctuated because the decomposition of the metastable SEI layer and the regeneration of a protective layer locally maintained some open-circuit voltage. The period from the ISC onset to venting was about 54 s in the baseline test.
Stage III — Safety-valve opening. Once the cell surface temperature reached approximately 219.13 °C, the internal pressure exceeded the rupture threshold of the safety valve. The valve opened noisily and a high-temperature gas mixture, mainly CO, H2, CO2 and CH4, was ejected. The voltage dropped almost to zero. A sustained flame appeared at the positive terminal in the unprotected test.
Stage IV — Thermal runaway. The cell continued to absorb heat until a thermal-runaway trigger temperature of 243.4 °C was reached. In the following seconds a large burning jet was discharged from the open end. The surface temperature rose from about 240 °C to 786.41 °C within approximately 15 s. The maximum temperature-rise rate reached 167.88 °C/s. After the combustible material was consumed, the temperature began to decrease and eventually returned to ambient. Key temperatures and times measured in the unprotected case are summarized in Table 3.
Table 3 — Characteristic points in the unprotected thermal-runaway test
| Event | Time (s) | Surface temperature (°C) |
|---|---|---|
| Onset of internal short circuit | 405 | 190.74 |
| Safety-valve opening | 459 | 219.13 |
| Thermal-runaway trigger | 494 | 243.40 |
| Maximum cell surface temperature | ≈ 509 | 786.41 |
One repetition of the unprotected thermal-runaway experiment produced a more violent result: the cell housing ruptured laterally and burning material was ejected through the side wall rather than only through the safety valve. This observation emphasizes the stochastic nature of the phenomena and illustrates why thermal-runaway propagation in a traction battery pack is difficult to predict deterministically: the flame direction, the moment of jet release and the fraction of energy released through the case can vary from one cell to another. A robust passive mitigation method must therefore be insensitive to the trigger location and to the particular characteristics of the failing cell.
The temperature-rise-rate history calculated from the measured surface data also helps to identify the stages. Until about 400 s, the rate of temperature rise remained below about 1 °C/s, typically between 0.5 and 0.9 °C/s. In the last few seconds before the runaway trigger, internal exothermic reactions became dominant; the heating rate quickly climbed from approximately 1.24 °C/s to the runaway peak value of 167.88 °C/s. After the burning jet disappeared, the cell surface began to cool rapidly. This sharp increase in temperature-rising rate is a clear early-warning indicator for imminent runaway in a traction battery pack if continuous temperature monitoring is available.
4. Single-Layer Insulation Experiments under Thermal Abuse
Ceramic-fibre silica aerogel composite boards were selected as the baseline passive thermal barrier. Their heat-transfer behaviour is much lower than that of conventional polymeric foams, and their maximum operating temperature is high enough to withstand the hot jet released during the first stage of thermal runaway. The material properties, as provided by the manufacturer, are given in Table 4. The boards had a size of 90 mm × 90 mm and were supplied in thicknesses of 1.5, 2, 3, 4 and 5 mm.
Table 4 — Properties of the ceramic-fibre silica aerogel composite
| Property | Value |
|---|---|
| Density (kg m−3) | 300 ± 20 |
| Thermal conductivity (W m−1 K−1) | 0.023 (25 °C); 0.030 (100 °C); 0.036 (300 °C); 0.072 (500 °C) |
| Specific heat capacity (J g−1 K−1) | 0.774 |
| Burning class | A, V0 |
| Hydrophobicity (%) | 98 |
The first exploratory experiments compared the thinnest available board, 1.5 mm, with the thickest board, 5 mm. The heating time was set to 1200 s; since thermal runaway of the unprotected cell occurred at 494 s, the chosen heating duration provided a safety factor of roughly 2.4 with respect to the unprotected trigger time.
4.1. Thermal-spread experiments with the 1.5 mm insulation layer
With a 1.5 mm aerogel layer placed between the heater and the cell, thermal runaway still occurred before the end of the 1200 s heating period. The sequence of visible events was similar to that observed for the unprotected cell, but the flame at safety-valve opening was absent; instead, grey-black smoke was expelled from the positive terminal. A burning jet with ejected fragments appeared during the final runaway stage. The measured temperatures at the top, middle and bottom of the cell showed broadly the same increasing trend before the runaway. After the onset of venting, the bottom temperature reached 779.36 °C, the top reached 725.06 °C and the middle reached 645.45 °C. These discrepancies demonstrate that the thermal runaway is initiated locally rather than uniformly over the entire jelly roll.
The characteristic trigger times of the 1.5 mm group are listed in Table 5. The protection delayed the occurrence of internal short circuit by about 200 s, delayed valve opening by about 278 s, and delayed thermal-runaway trigger by approximately 357 s. Thus the delay time increased by about 72.3% compared with the unprotected case. The insulation board also altered the temperature-rise rate before the runaway: during the first 200 s the effect was negligible, but between 200 s and the trigger point the 1.5 mm aerogel reduced the heating rate, keeping it below 0.3 °C/s for a substantial portion of the heating phase.
Table 5 — Characteristic events with a 1.5 mm single-layer insulation
| Event | Time (s) | Surface temperature (°C) |
|---|---|---|
| Onset of internal short circuit | 605 | 194.20 |
| Safety-valve opening | 737 | 218.08 |
| Thermal-runaway trigger | 851 | 231.75 |
4.2. Thermal-spread experiments with the 5 mm insulation layer
When the insulation thickness was increased to 5 mm, the cell did not undergo thermal runaway during the 1200 s heating period. No internal short circuit was detected from the voltage trace, and the voltage remained almost constant with only a small drift of approximately 0.01 V. The surface temperature rose monotonically, with a maximum measured rise rate of about 0.403 °C/s. The average temperature increase rate was much smaller than that with the 1.5 mm thick board. After approximately 700 s the rate of temperature increase slowed progressively, suggesting that a quasi-steady thermal state was approached. These findings confirm that a 5 mm thick ceramic-fibre silica aerogel board is able to stop thermal propagation in the tested configuration.
The comparison of the 1.5 mm and 5 mm insulation groups clearly demonstrates how strongly the barrier thickness affects the thermal protection capability in a traction battery pack. However, thicker insulation also consumes valuable volume and adds mass. Therefore, a systematic optimization of the insulation-layer thickness was necessary; this was accomplished through numerical simulation after validating the method against the 5 mm experimental data.
5. Numerical Study of the Single-Layer Insulation Structure
The simulations were performed with the commercial CFD solver STAR-CCM+. The computational domain replicated the constant-temperature chamber with dimensions of 0.5 m × 0.5 m × 0.5 m. The heater, the aerogel board and the cylindrical cell were modelled as solid bodies. Air around them was treated as an ideal gas with temperature-dependent properties. The Realizable \(k\)-\(\varepsilon\) turbulence model was used to resolve the buoyant natural-convection flow. The gravitational acceleration was set to 9.81 m/s2. The initial temperature of the whole domain was 27 °C. A heat source of 350 W was applied uniformly within the heated section of the rod. Polyhedral meshes were generated with ten prism boundary layers near the solid walls; the thickness of the prism layer was 0.2 mm and the stretching factor was 1.2.
The governing equations solved by the code are the continuity equation, the momentum equation and the energy equation. For the present low-Mach number buoyant flow, they can be written symbolically as
$$ \frac{\partial \rho}{\partial t} + \nabla \cdot (\rho \mathbf{u}) = 0 \qquad\qquad (6) $$
$$ \frac{\partial (\rho \mathbf{u})}{\partial t} + \nabla \cdot (\rho \mathbf{u} \mathbf{u}) = -\nabla p + \nabla \cdot \boldsymbol{\tau} + \rho \mathbf{g} \qquad\qquad (7) $$
$$ \frac{\partial (\rho c_p T)}{\partial t} + \nabla \cdot (\rho c_p \mathbf{u} T) = \nabla \cdot (k \nabla T) + \dot{q}”’ \qquad\qquad (8) $$
in which \(\mathbf{u}\) is the velocity vector, \(p\) the pressure, \(\boldsymbol{\tau}\) the viscous stress tensor, \(c_p\) the specific heat at constant pressure and \(\dot{q}”’\) the volumetric heat source. The latter is zero inside the aerogel and the cell because chemical heat generation is neglected in this study.
Before selecting the optimum aerogel thickness, a grid-independence study was performed using mesh counts from 0.86 million to 8.28 million. Monitored temperatures at three points along the cell circumference and height changed by less than 0.4‰ when the mesh count was increased beyond 3.04 million. To balance accuracy and computational cost, a mesh containing approximately 5.59 million cells was used for the final simulations. Table 6 reports the monitored temperatures for all investigated mesh levels.
Table 6 — Grid-independence study results
| Case | Mesh count | T1 (°C) | T2 (°C) | T3 (°C) |
|---|---|---|---|---|
| 1 | 855246 | 282.397 | 181.959 | 160.865 |
| 2 | 1134693 | 282.134 | 181.172 | 161.151 |
| 3 | 1603588 | 282.324 | 180.736 | 161.628 |
| 4 | 2029072 | 282.353 | 180.933 | 161.723 |
| 5 | 3041836 | 282.455 | 181.083 | 161.944 |
| 6 | 5586296 | 282.558 | 181.042 | 161.981 |
| 7 | 8275925 | 282.564 | 181.055 | 161.894 |
The computed temperature curves for the 5 mm single-layer configuration were compared with the measured experimental data at the same thermocouple positions. The maximum deviation between simulation and experiment was approximately 5.6%, which confirmed that the numerical method correctly reproduced the overall heat-transfer process. The shape of the temperature curves, including their steadily decreasing slope with time, was reproduced accurately by the simulation.
6. Optimum Single-Layer Thickness Selection
Since the aerogel manufacturer supplied boards only in discrete thicknesses, a curve-fitting approach was used to determine the best available single-layer thickness. Twelve simulations were carried out for aerogel thicknesses starting at 5 mm and decreasing in steps of 0.3 mm down to 1.7 mm. The maximum battery-surface temperature at a given thickness corresponded to the average value of the three monitoring points after the full 1200 s exposure. The results are listed in Table 7.
Table 7 — Predicted maximum battery-surface temperature for various aerogel thicknesses
| Insulation thickness (mm) | Maximum surface temperature (°C) |
|---|---|
| 5.0 | 133.60 |
| 4.7 | 135.60 |
| 4.4 | 137.90 |
| 4.1 | 140.20 |
| 3.8 | 144.00 |
| 3.5 | 149.60 |
| 3.2 | 155.50 |
| 2.9 | 166.10 |
| 2.6 | 185.00 |
| 2.3 | 204.90 |
| 2.0 | 225.68 |
| 1.7 | 247.80 |
Fitting these twelve points to a cubic polynomial yielded the following relation between the temperature \(T_{max}\) and the aerogel thickness \(x\):
$$ T_{max}(x) = -2.445 x^3 + 38.87 x^2 – 206.4 x + 500.8 \qquad\qquad (9) $$
with a coefficient of determination \(R^2 = 0.9975\). The maximum acceptable cell-surface temperature was defined conservatively as 142.1 °C, corresponding to the surface temperature measured 100 s before internal short circuit in the unprotected test. A cell surface temperature below this value ensures that the cell does not approach its internal-short-circuit threshold. Among the available discrete thicknesses, 3 mm would give a predicted temperature of roughly 165 °C, which is too high; 4 mm gives about 140.6 °C, satisfying the safety margin. Therefore, the 4 mm aerogel was selected as the best single-layer insulation thickness for further study.
When a dedicated simulation was run for the 4 mm thick aerogel board, the average monitoring-point temperature at 1200 s was 137.82 °C, in very good agreement with the value predicted from the fitted curve. Subsequent experimental validation with the 4 mm board confirmed that the cell surface temperature remained slightly below the safety threshold. The measured voltage trace did not show any evidence of internal short circuit. The experimental and numerical temperatures agreed within a maximum deviation of about 5.6% throughout the heating period, providing a robust verification of the thickness-selection methodology.
An additional benefit of the numerical simulation is the possibility of extracting details that are difficult to measure in a highly transient thermal-abuse experiment. In the final state of the 4 mm protection simulation, the temperature on the cell-facing side of the aerogel was much lower than on the heater-facing side, and the radial temperature gradient inside the cell was pronounced. The horizontal thermal asymmetry of the cell resulted in a surface temperature difference of more than 70 °C between the heated side and the shielded side. This result highlights the challenge of thermal-management design for a traction battery pack: even when global protection is successful, local thermal gradients can be large enough to generate thermomechanical stress in neighbouring components.
The simulation also provided the velocity field of the air surrounding the heater and the cell. Natural convection developed as the heated air rose along the heater and the hot side of the aerogel, while colder air descended near the shielded side of the cell. The maximum instantaneous air velocity in the computational domain at 1200 s was about 0.64 m/s. Most of the high-velocity fluid was located above the heater, where the air temperature was highest and the buoyancy force strongest. To quantify the effect of the aerogel barrier on convective cooling, the air velocity in the horizontal mid-plane was extracted at several times. After 200 s of heating, the maximum velocity near the heater side was approximately 0.4 m/s, whereas the maximum velocity on the battery side was only about 0.14 m/s, i.e., approximately 35% of the heater-side value. This reduction in air velocity near the cell comes from the lower temperature of the cell-side air and is a further manifestation of the radiation and conduction blocking action of the silica aerogel.
7. Module-Level Simulation of a Three-by-Three Traction Battery Pack
To demonstrate the practicality of the single-layer 4 mm aerogel insulation, the same barrier was then applied in a 3×3 cell array representing a small sectional module of a traction battery pack. Each cell was separated from its neighbours by the 4 mm ceramic-fibre silica aerogel board. A heating rod was placed at three different locations: at the center of the module, at one diagonal corner, and at the outer side of the module. The position of the rod was varied to reproduce different possible failure origins inside a battery pack. The maximum cell-surface temperature of the monitored cells is reported in Table 8 for the three arrangements.
Table 8 — Module-level simulation results after 1200 s
| Heater location | Cell identifier | Maximum surface temperature (°C) |
|---|---|---|
| Case A (center) | Cell 1 | 60.18 |
| Case A (center) | Cell 2 | 133.55 |
| Case B (diagonal corner) | Cell 1 | 28.30 |
| Case B (diagonal corner) | Cell 4 | 30.65 |
| Case B (diagonal corner) | Cell 5 | 59.98 |
| Case B (diagonal corner) | Cell 7 | 35.17 |
| Case B (diagonal corner) | Cell 8 | 133.73 |
| Case C (outer side) | Cell 1 | 134.31 |
| Case C (outer side) | Cell 3 | 60.08 |
| Case C (outer side) | Cell 4 | 132.65 |
| Case C (outer side) | Cell 6 | 30.69 |
| Case C (outer side) | Cell 7 | 34.69 |
Even in the most severe cases, the battery surface temperature did not exceed 134.31 °C, remaining below the safety threshold adopted in the single-cell study. The cell located farthest from the heating rod reached a maximum surface temperature of only about 30 °C. This nearly negligible temperature increase demonstrates the excellent thermal isolation provided by the silica aerogel layer. The flow field inside the 3×3 module was qualitatively different from that of the single-cell arrangement because the inter-cell gaps created narrow vertical channels in which buoyancy-driven flow was accelerated. Nevertheless, the aerogel kept the cell-side air temperature low, and the peak air speed remained modest. The temperature-field plots confirmed that, on a horizontal plane, the major temperature gradient appeared across the aerogel board itself rather than across the cells.
8. Optimized Three-Layer Insulation Structure
Although the 4 mm single-layer aerogel board satisfied the safety requirement, it still allowed the cell surface to approach the thermal-runaway boundary. To increase the safety margin without significantly increasing the total thickness, the aerogel board was replaced by a three-layer laminate composed of the 4 mm aerogel sheet and two thin metal sheets bonded to its faces. The metal sheets had the same 90 mm × 90 mm footprint as the aerogel. The heat-blocking mechanism of this laminate is twofold. First, the metal skin facing the heater has a very low surface emissivity, and therefore reflects a large fraction of the incoming thermal radiation back toward the heat source. Second, the aerogel core still provides excellent conduction blocking; the second metal skin improves the uniformity of the temperature on the protected side and also blocks radiative exchange between the aerogel and the cell.
To quantify the effect of the metal-layer thickness, copper sheets with thicknesses of 0.1, 0.2, 0.3 and 0.4 mm were considered, together with an aerogel-only control sample. The temperature on the back side of the insulation structure, i.e., the side facing the protected cell, was measured at three vertical positions with thermocouples. The measured temperature curves show that the heater surface quickly reached a quasi-steady temperature of about 632 °C after 500 s. All four metal-clad configurations produced a back-side temperature below 160 °C. The 0.2, 0.3 and 0.4 mm copper sheets gave very similar results, with a difference of only about 3 °C after the steady state was reached. The 0.1 mm copper sheet was slightly less effective, with an average back-side temperature roughly 10 °C higher than that of the 0.2 mm sample. This is explained by oxidation: the very thin copper sheet oxidized rapidly at high temperature, and the emissivity of the oxidized copper surface increased to approximately 0.6–0.7, reducing its ability to reflect radiation. Since the 0.2 mm Cu sheet gave essentially the same performance as the thicker sheets, the 0.2 mm thickness was chosen as the baseline for the subsequent investigation.
A comparison between the copper–aerogel laminate and the bare aerogel board is shown in Table 9. The bare aerogel reached a steady back-side temperature of roughly 309 °C, whereas the three-layer copper-laminated board reached only about 150 °C. The average temperature-rise rate during the period of quasi-linear increase was reduced from about 1.11 °C/s for the bare aerogel to about 0.255 °C/s for the copper–aerogel laminate.
Table 9 — Performance comparison of the single-layer aerogel and the copper-laminated three-layer structure
| Property | Single-layer aerogel | Cu–aerogel three-layer structure |
|---|---|---|
| Steady back-side temperature (°C) | 309.13 | 149.62 |
| Average temperature rise rate (°C/s) | 1.11 | 0.255 |
| Time to quasi-steady state (s) | ≈ 450 | ≈ 800 |
Since the metal skin material plays an important role in the radiation reflection and oxidation resistance, four metals were compared experimentally using a sheet thickness of 0.2 mm: copper, tungsten, iron and Inconel 718. The back-side temperature responses are summarized in Table 10. Copper and tungsten produced the lowest steady-state temperature, while iron produced the highest. Inconel 718 reached thermal equilibrium sooner than the other materials but the steady temperature was slightly higher than that of copper. The difference between copper and tungsten was small, but copper had a slightly lower maximum temperature and a shorter time to reach steady state.
Table 10 — Comparison of metal skin materials
| Metal | Average rise rate from 200 to 400 s (°C/s) | Average temperature from 1100 to 1200 s (°C) | Maximum temperature (°C) | Time to quasi-steady state (s) |
|---|---|---|---|---|
| Copper | 0.251 | 151.41 | 152.07 | 660 |
| Tungsten | 0.252 | 151.64 | 152.88 | 881 |
| Iron | 0.290 | 162.56 | 163.56 | 707 |
| Inconel 718 | 0.335 | 154.35 | 157.85 | 564 |
The four metal-clad insulation structures were then subjected to actual thermal-spread control tests, in which a fully charged cell was placed behind the laminate and heated for 1200 s. The voltage trace remained stable in all configurations, confirming that no internal short circuit occurred. The maximum cell-surface temperatures are listed in Table 11. The copper laminate gave the best result: the cell surface reached only 91.59 °C at the end of the test. Tungsten, iron and Inconel 718 produced maximum cell temperatures between 98 and 101 °C.
Table 11 — Cell surface temperature after 1200 s heating with different metal skins
| Metal skin | Maximum cell-surface temperature (°C) |
|---|---|
| Copper | 91.59 |
| Tungsten | 98.30 |
| Iron | 101.24 |
| Inconel 718 | 100.53 |
Compared with the single-layer aerogel insulation structure, whose protective effect during the battery test produced a surface maximum of approximately 130.77 °C, the optimized 0.2 mm copper–aerogel three-layer structure lowered the cell-surface temperature by 39.18 °C, corresponding to a 29.96% reduction. Such an improvement is significant for the safety margin of a traction battery pack, since it reduces the risk of triggering separator melting even in the presence of considerable external heat flux.
The main trade-off of the three-layer structure is its mass increase. For a 90 mm × 90 mm area, the two 0.2 mm copper sheets weigh about 29.03 g, whereas the 90 mm × 90 mm × 4 mm aerogel board weighs only about 9.72 g. The total mass of the laminate is therefore 38.75 g, an increase of approximately 298.7% compared with the bare aerogel. From a lightweight design perspective, future improvements should aim at selecting a metal skin with low emissivity, high oxidation resistance and low density. If a thin, highly oxidation-resistant metal foil can be developed, the thickness of the metal layer could also be reduced below 0.2 mm while preserving its radiative heat-reflection effect.
9. Heat-Blocking Mechanism of the Three-Layer Structure
The excellent performance of the ceramic-fibre silica aerogel used here is due to the multi-scale pore structure of its nanostructured network. The material combines ceramic fibres with silica aerogel particles, resulting in very low solid conduction because the particle size approaches the phonon mean free path. The tortuous path through the nanoporous network further suppresses gas-phase heat transfer. In addition, the small pore dimensions inhibit local gas convection. At elevated temperature, radiation transfer through the aerogel becomes more important, which explains why the apparent thermal conductivity of the composite increases slightly with temperature.
When a metal sheet is attached to the heater-facing side of the aerogel, the sequence of heat-transfer events changes decisively. The polished copper surface has an emissivity of approximately 0.03, compared with about 0.794 for the silica aerogel surface. According to Kirchhoff’s law, a low-emissivity surface also has a low absorptivity; consequently, the copper skin reflects most of the incident infrared radiation back toward the heat source. The small absorbed fraction is then conducted through the aerogel core, whose low thermal conductivity creates a large temperature drop. The second metal skin, located on the cell-facing side, serves as a radiative shield that prevents the aerogel from radiating directly to the cell. It also equalizes the temperature distribution on the protected surface.
Another contribution is the reduction of convective heat transfer. Because the metal-clad structure limits the temperature of the air adjacent to the protected cell, the buoyancy-driven velocity near the cell remains low. This reduces the convective heat-transfer coefficient on the cell surface and therefore lowers the heat flux entering the cell. The combined effect of radiation reflection, conduction blocking and convection suppression explains why the three-layer structure significantly outperforms the bare aerogel board of identical core thickness. The design can be easily transferred to the module level of a traction battery pack, where thin metal-faced aerogel sheets can be placed between every pair of neighbouring cells to act as both thermal barriers and mechanical separators.
10. Conclusions
The experimental and numerical work described here has demonstrated a clear design route for passive thermal-runaway propagation control in a vehicle traction battery pack. The following main conclusions were obtained:
1. Under thermal abuse supplied by a 350 W heating rod, a fully charged 18650 NCA cell experienced a reproducible four-stage runaway process. The internal-short-circuit temperature was about 190.74 °C, the safety-valve opening temperature was about 219.13 °C and the thermal-runaway trigger temperature was about 243.4 °C. The maximum surface temperature during runaway reached 786.41 °C with a peak temperature-rise rate of 167.88 °C/s.
2. A 1.5 mm ceramic-fibre silica aerogel board significantly delayed thermal runaway but could not prevent it. The 5 mm board suppressed thermal runaway throughout the 1200 s heating period. Simulation results were used to interpolate the optimum single-layer thickness at 4 mm, satisfying the 142.1 °C safety criterion while minimizing the insulation-layer volume.
3. Module-level simulations with the 4 mm single-layer insulation placed in a 3×3 battery module showed that the maximum cell-surface temperature remained below 134.31 °C for three possible heater locations. This confirms that the selected single-layer barrier is adequate for practical module implementation in a traction battery pack.
4. A three-layer laminate, formed by bonding two 0.2 mm copper sheets onto the 4 mm aerogel board, lowered the maximum protected-cell surface temperature by 39.18 °C, i.e., by 29.96%, compared with the bare aerogel board. The metal skins reflect a large portion of the thermal radiation and reduce convective heating of the protected cell. Copper was found to be the most effective tested metal layer, followed by tungsten, Inconel 718 and iron.
Future work will extend the concept to full-scale module fire tests, investigate the performance of thinner metal foils with enhanced oxidation resistance, and incorporate coupled electrochemical-thermal models to describe more accurately the behaviour of the cell during the early stages of internal heating. The ultimate goal is the development of a lightweight, compact and reliable thermal-propagation barrier that can be integrated directly into the pack architecture of a traction battery pack without degrading its energy density.