As a researcher specializing in vehicle body engineering and battery safety, I have devoted my recent work to a very practical yet destructive challenge: the suppression of thermal runaway propagation in EV battery pack systems. My investigations center on 18650-type lithium-ion power cells with a nickel–cobalt–aluminum oxide (NCA) cathode. Such cylindrical cells are widely adopted in electric vehicles because of their high energy density and favourable cycle performance. However, the same high energy density, when combined with a triggering event such as mechanical abuse, electrical abuse, or thermal abuse, can lead to catastrophic thermal runaway. Once thermal runaway occurs in one cell, the enormous amount of heat released within a short period often propagates to neighbouring cells, causing a chain reaction that can result in fire, explosion, and the complete loss of an EV battery pack.
To mitigate this hazard, I have studied passive control strategies based on thermal barrier materials. A passive approach is particularly attractive for an EV battery pack because it requires no external energy input, consumes no active cooling fluid, and can be integrated directly between individual battery cells. The core material I have selected is a ceramic-fiber silica-aerogel composite, which offers extremely low thermal conductivity and excellent fire resistance. In this article, I present the main results of my experimental and numerical investigations, with an emphasis on the performance of single-layer and three-layer insulation structures in controlling thermal spread within an EV battery pack under thermal abuse conditions.
Research Background and Objectives
The popularity of electric vehicles continues to grow due to the combined pressure of global climate change and the depletion of conventional energy resources. Many studies have pointed out that lithium-ion batteries will remain the dominant energy storage technology for electric vehicles in the near future. Among the various lithium-ion chemistries, NCA-based cathode materials provide high specific energy and thus enable longer driving ranges. Nevertheless, the thermal stability of NCA cells is lower than that of lithium iron phosphate cells, which implies a higher safety risk when an EV battery pack is subjected to abusive conditions. During an internal short circuit or an externally induced overheating event, multiple exothermic decomposition reactions occur inside the cell. These reactions include the decomposition of the solid electrolyte interphase (SEI), the reaction between intercalated lithium and the electrolyte, the decomposition of the cathode material, and the further decomposition of the electrolyte solvent. The heat released by these reactions may lead to a rapid self-heating process, eventually culminating in jet flames, explosion, and fire.
In my research, I applied a high-temperature heating rod as the thermal abuse source to mimic real-world scenarios in which a cell is locally overheated inside an EV battery pack. The entire thermal runaway process was analysed using battery surface temperature and voltage data. The key objective was to identify the optimal thickness of a single-layer silica-aerogel insulation structure that can block thermal spread while keeping the added volume and mass to a minimum. In addition, I investigated an optimised three-layer insulation structure, in which thin metallic sheets are attached to both sides of the aerogel board, and evaluated how the metal thickness and material affect the passive thermal-barrier performance.
Theoretical Foundations
Heat transfer in the experimental setup involves all three fundamental mechanisms: conduction, convection, and radiation. Fourier’s law describes the conductive heat flux in a solid material:
$$q = -\lambda \frac{dT}{dx}$$
where $\lambda$ is the thermal conductivity and $dT/dx$ is the temperature gradient in the direction of heat flow. The convective heat transfer between the heated surfaces and the surrounding air can be described by Newton’s law of cooling:
$$\Phi = h A \Delta T$$
where $h$ is the surface heat transfer coefficient, $A$ is the surface area, and $\Delta T$ is the temperature difference between the surface and the fluid.
Thermal radiation is especially important at elevated temperatures. The radiative heat emitted by a blackbody is given by the Stefan–Boltzmann equation:
$$\Phi = \varepsilon A \sigma T^4$$
where $\varepsilon$ is the emissivity of the surface, $\sigma$ is the Stefan–Boltzmann constant, and $T$ is the absolute surface temperature. For the thermal runaway process itself, lithium-ion cell decomposition reactions can be represented by Arrhenius-type kinetic laws. As an example, the SEI decomposition rate can be written as:
$$\frac{\mathrm{d}\chi_{\mathrm{SEI}}}{\mathrm{d}t} = -A_{\mathrm{SEI}}\,\chi_{\mathrm{SEI}} \exp\left(-\frac{E_{a,\mathrm{SEI}}}{RT}\right)$$
The heat generation rate associated with this reaction is related to the reaction rate by:
$$\dot{Q}_{\mathrm{SEI}} = -m_{\mathrm{SEI}}\, H_{\mathrm{SEI}}\, \frac{\mathrm{d}\chi_{\mathrm{SEI}}}{\mathrm{d}t}$$
where $m_{\mathrm{SEI}}$ is the effective mass of the SEI layer and $H_{\mathrm{SEI}}$ is the specific reaction enthalpy. Similar formulations apply to the negative-electrode reaction, the positive-electrode decomposition, and the electrolyte decomposition. The total heat release rate during thermal runaway is therefore the sum of all such contributions.
Experimental Configuration
Cell Specifications
The test cells were commercial Sony US18650VTC6 lithium-ion batteries. The positive electrode used a nickel-rich NCA material, whereas the negative electrode was graphite based. Table 1 summarises the main cell parameters.
| Parameter | Value |
|---|---|
| Diameter × Height / mm | 18 × 65 |
| Nominal capacity / mAh | 3000 |
| Nominal voltage / V | 3.6 |
| Charge cut-off voltage / V | 4.2 |
| Discharge cut-off voltage / V | 2.0 |
| Operating temperature / °C | 0–60 |
| Mass / g | 46.6 ± 1.5 |
The passive thermal barrier material used throughout the experiments was a ceramic-fibre silica-aerogel composite board supplied by a manufacturer. The material provides a very low thermal conductivity at room temperature, strong hydrophobicity, and a UL-94 V0 flame rating. Table 2 lists the material properties provided by the manufacturer.
| Property | Value |
|---|---|
| Density / kg m⁻³ | 300 ± 20 |
| Thermal conductivity @ 25 °C / W m⁻¹ K⁻¹ | 0.023 |
| Thermal conductivity @ 100 °C / W m⁻¹ K⁻¹ | 0.030 |
| Thermal conductivity @ 300 °C / W m⁻¹ K⁻¹ | 0.036 |
| Thermal conductivity @ 500 °C / W m⁻¹ K⁻¹ | 0.072 |
| Specific heat capacity / J g⁻¹ K⁻¹ | 0.774 |
| Combustion rating | Class A, V0 |
| Hydrophobicity / % | 98 |
Test Platform and Instrumentation
All high-temperature tests were performed inside a temperature-controlled chamber with internal dimensions of 0.5 m × 0.5 m × 0.5 m. The chamber was set to 27 °C for every test. A heating rod with a length of 90 mm in the heating zone and a diameter of 16 mm was used as the external heat source. The heater was powered by a DC power controller with an output power set to 350 W. Battery surface temperatures were measured with K-type thin-patch thermocouples connected to a data acquisition instrument. Cell voltage was monitored using a high-precision digital multimeter. An infrared thermal camera recorded the thermal runaway events through the observation window of the chamber.
The cells were clamped vertically on an adjustable platform in the centre of the chamber. This allowed precise positioning of the heating rod, the insulation board, and the battery relative to one another. In the thermal-runaway test, the battery and heating rod were placed in direct horizontal contact. In the thermal-spread control tests, the aerogel board was inserted between the battery and the heating rod. The three thermocouples were attached on the battery surface close to the heated side at the upper, middle, and lower positions, respectively. The average of these three measurements was used as the battery-side surface temperature.
Single-Cell Thermal Runaway Experiment
Experimental Description
All cells were fully charged to 100% SOC before each experiment. The charging procedure consisted of a constant-current charge at 0.5 C until the voltage reached 4.2 V, followed by a constant-voltage charge until the current dropped below 150 mA. After charging, the cells were allowed to rest for one hour at room temperature. In the thermal-runaway test, the heating rod remained in direct contact with the cell surface. The heating rod was powered continuously until thermal runaway occurred. The key operating parameters are given in Table 3.
| Parameter | Value |
|---|---|
| Battery SOC | 100% (3000 mAh) |
| Heating power / W | 350 |
| Heater–battery horizontal gap / mm | 0 |
| Heating zone length / mm | 90 |
| Battery bottom height above chamber floor / mm | 100 |
| Heater bottom height above chamber floor / mm | 22.5 |
Four-Stage Process Analysis
The measured temperature and voltage curves for a representative single-cell thermal runaway experiment are shown in the temperature–time and voltage–time curves that I obtained. The entire process can be divided into four stages. In the first stage, the cell surface temperature rises continuously under the external heating. Minor voltage fluctuation is observed, and the cell remains macroscopically intact. In the second stage, the internal separator begins to melt and a partial internal short circuit develops. The measured cell voltage drops sharply, but it does not immediately fall to zero. Instead, the voltage oscillates because of successive SEI decomposition and healing events as well as the non-uniform melting of the separator. In the third stage, the safety valve opens and high-temperature gas is released. The voltage then collapses to near zero, and a sustained flame can appear at the positive terminal. In the fourth stage, thermal runaway itself occurs with a violent jet of burning material from the cell. The surface temperature rises rapidly and then decays.
Table 4 summarises the characteristic temperatures measured in the single-cell tests.
| Characteristic Event | Time / s | Temperature / °C |
|---|---|---|
| Triggering of internal short circuit | 405 | 190.74 |
| Safety valve opening | 459 | 219.13 |
| Onset of thermal runaway | 494 | 243.40 |
| Maximum surface temperature | 510 | 786.41 |
The maximum rate of temperature rise reached 167.88 °C/s during the final runaway stage. Between the first detection of an internal short circuit and the onset of thermal runaway only about 89 s elapsed. This extremely short time window demonstrates how challenging it is to interrupt a fully developed thermal runaway event. On the other hand, it also highlights the importance of preventing the early heat accumulation from ever reaching a critical state.
The visual records captured by the infrared camera confirm several important phenomena: the opening of the safety valve, the ejection of high-temperature vapor, a brief stable flame above the positive terminal, and the abrupt release of a burning jet. In one additional test, I observed side-wall rupture with an explosion-like release of material. This indicates that thermal runaway does not always follow the same failure path. The stochastic nature of the cell failure mode underscores the need for robust passive barriers inside an EV battery pack. If a cell ruptures laterally, the abusing heat flux can directly attack neighbouring cells through a much shorter path.
Temperature Rise Rate Analysis
After processing the temperature data, I computed the instantaneous temperature rise rate for each second. Before thermal runaway, the surface temperature rise rate remained below 1 °C/s, and for most of the heating period it was about 0.5 °C/s. A gradual increase from 0.5 °C/s to about 0.9 °C/s occurred between the internal short circuit at 405 s and the onset of thermal runaway at 494 s. The sudden jump in the rise rate to the maximum value of 167.88 °C/s coincided with the final runaway stage. These observations show that during the external-heating phase the cell temperature increase is governed largely by the external heat input rather than by internal chemical reactions. It is only after the separator fails that the internal exothermic reactions become the dominant heat source.
Preliminary Experimental Study of Thermal-Spread Control
Thickness Selection and Safety Criterion
Before comparing different aerogel thicknesses, I set a test duration of 1200 s as the pass-fail cut-off. This duration was chosen with a safety coefficient of about 2.4 relative to the 494 s heating time required for thermal runaway in the unprotected cell. Thus, if no thermal runaway is observed within 1200 s of continuous heating at the rated heater power, the tested barrier is considered effective. Initially, I selected the thinnest available board (1.5 mm) and the thickest board (5 mm) for physical testing.
Results with 1.5 mm Aerogel
With a 1.5 mm aerogel layer placed between the heater and the battery, thermal runaway still occurred before 1200 s. However, the thermal runaway characteristics changed. The failure sequence still included internal short circuit, safety valve opening, and thermal runaway, but no ignited flame was observed immediately at the venting stage. Instead, grey-black smoke was emitted from the positive terminal, followed by a brief violent release of fragmented materials. Important time and temperature data are listed in Table 5.
| Event | Time / s | Temperature / °C |
|---|---|---|
| Internal short circuit | 605 | 194.2 |
| Safety valve opening | 737 | 218.08 |
| Thermal runaway | 851 | 231.75 |
| Maximum surface temperature | 868 | 779.36 |
Compared with the unprotected cell, the 1.5 mm aerogel extended the time to thermal runaway by 357 s, which corresponds to an increase of about 72.3%. The characteristic temperatures did not change substantially, indicating that the thin aerogel layer influences the heat transfer rate without changing the intrinsic thermal stability of the active materials. The average temperature rise rate after 200 s was noticeably reduced, remaining below 0.3 °C/s for most of the heating process before runaway. However, the protection was insufficient because the total heat input still drove the cell into the self-heating regime.
Results with 5 mm Aerogel
When the aerogel thickness was increased to 5 mm, the battery did not undergo thermal runaway within the entire 1200 s heating duration. The cell voltage remained essentially stable, with a variation of only about 0.01 V. No internal short circuit was observed. The surface temperature rose continuously but reached only about 120 °C at the end of the experiment. These results demonstrate that a sufficiently thick layer of silica aerogel can successfully suppress thermal propagation under the given thermal abuse conditions. The maximum heating rate measured during the entire experiment was only about 0.403 °C/s.
To further compare the thermal insulation effect, I calculated the average temperature rise rates at intervals of 50 s and 100 s for both the 1.5 mm experiments and the 5 mm experiments. In both cases, the temperature rise rate initially increased and then decreased. The 5 mm board consistently produced a smaller average rise rate than the 1.5 mm board. This behaviour is typical of silica-aerogel composite materials, in which the thermal conductivity changes with temperature and time.
Numerical Simulation of the Single-Layer Insulation Structure
Simulation Setup
To investigate intermediate thickness values without manufacturing a large number of physical samples, I constructed a computational fluid dynamics model based on the finite volume method. The computational domain replicated the physical test chamber, with the heating rod, silica-aerogel board, and lithium-ion cell positioned in the same relative arrangement as in the experiments. The air was modelled as an ideal gas with natural convection. I selected the Realizable k-epsilon turbulence model for the flow field. The cell was treated as a solid with anisotropic thermal conductivity: 0.413 W m⁻¹ K⁻¹ in the radial direction and 1.780 W m⁻¹ K⁻¹ in the axial direction. The volumetric heat capacity and density of the cell were set according to values reported in the literature for 18650 NCA cells. The relevant boundary conditions and material properties are listed in Table 6.
| Category | Parameter | Value |
|---|---|---|
| Air | Density / kg m⁻³ | 1.18415 |
| Specific heat / J kg⁻¹ K⁻¹ | 1003.62 | |
| Thermal conductivity / W m⁻¹ K⁻¹ | 0.026 | |
| Dynamic viscosity / Pa s | 1.85508 × 10⁻⁵ | |
| Heating rod | Size (diameter × height) / mm | 16 × 90 |
| Heating power / W | 350 | |
| Surface emissivity | 0.2 | |
| Silica aerogel | Size (L × W × H) / mm | 90 × 90 × 5 |
| Density / kg m⁻³ | 300 | |
| Specific heat / J kg⁻¹ K⁻¹ | 774 | |
| Surface emissivity | 0.794 | |
| Battery | Size (diameter × height) / mm | 18 × 65 |
| Density / kg m⁻³ | 2817.412 | |
| Specific heat / J kg⁻¹ K⁻¹ | 958.5 | |
| Radial thermal conductivity / W m⁻¹ K⁻¹ | 0.413 | |
| Axial thermal conductivity / W m⁻¹ K⁻¹ | 1.780 | |
| Surface emissivity | 0.9 |
The computational mesh consisted of polyhedral cells with additional boundary-layer prisms near the solid walls. A grid independence study was performed at seven mesh densities from roughly 0.85 million to 8.28 million cells. Three monitor points on the mid-height battery surface were used for comparison. The difference in computed temperature between the mesh with about 3.0 million cells and the mesh with about 8.3 million cells was less than 0.4‰, so the final simulations were carried out with a mesh of about 5.6 million cells to balance accuracy and computational cost. The time step was set to 0.1 s, with 5 iterations per time step, covering a total simulation time of 1200 s.
Validation against Experiment
To validate the simulation model, I first reproduced the 5 mm aerogel thermal-spread control experiment. The simulated temperature at the three battery surface monitoring points were averaged and compared with the experimentally recorded values. The overall temperature-rise trend matched well. The maximum deviation between simulation and experiment was approximately 5.6%, which is acceptable for the subsequent parametric studies. This validation indicated that the numerical model correctly captures the dominant heat transfer pathways, including conduction through the aerogel board, radiation from the heated rod, and natural convection in the air gap. I then used the model to explore the relationship between aerogel thickness and the resulting battery-surface temperature after 1200 s of heating.
Optimisation of the Single-Layer Aerogel Thickness
Simulation Result Matrix
I simulated twelve aerogel thicknesses ranging from 1.7 mm to 5.0 mm. For each case, the maximum battery-surface temperature at the monitoring points is listed in Table 7.
| Case No. | Aerogel thickness / mm | Max battery surface temperature / °C |
|---|---|---|
| 1 | 5.0 | 133.60 |
| 2 | 4.7 | 135.60 |
| 3 | 4.4 | 137.90 |
| 4 | 4.1 | 140.20 |
| 5 | 3.8 | 144.00 |
| 6 | 3.5 | 149.60 |
| 7 | 3.2 | 155.50 |
| 8 | 2.9 | 166.10 |
| 9 | 2.6 | 185.00 |
| 10 | 2.3 | 204.90 |
| 11 | 2.0 | 225.68 |
| 12 | 1.7 | 247.80 |
I fitted a cubic polynomial to the thickness-temperature data using a least-squares procedure. The fitted equation is:
$$T_{\mathrm{max}}(x) = -2.445 x^3 + 38.87 x^2 – 206.4 x + 500.8$$
where $x$ is the aerogel thickness in millimetres and $T_{\mathrm{max}}$ is the maximum battery surface temperature at 1200 s. The coefficient of determination for the fit is R² = 0.9975, indicating an excellent correlation. When the five manufacturer-supplied thicknesses are substituted into the fitted equation, the corresponding temperatures are presented in Table 8.
| Aerogel thickness / mm | Predicted max battery surface temperature / °C |
|---|---|
| 5.0 | 134.925 |
| 4.0 | 140.640 |
| 3.0 | 165.415 |
| 2.0 | 223.920 |
| 1.5 | 270.406 |
Selection of 4 mm as Optimal Thickness
I selected the optimal thickness based on two criteria. First, the cell must remain safe for the entire 1200 s test. Second, the added thickness and mass must be as small as possible. In the unprotected thermal-runaway test, the internal short circuit occurred at a surface temperature of 190.74 °C. Since internal short circuits are already associated with substantial internal heat generation, I used a more conservative safe threshold: the battery surface temperature should be kept below the internal-short-circuit temperature minus a margin. The temperature at 100 s before internal short-circuit onset in the unprotected test was about 142.1 °C. Using this value as the safety criterion, both the 4 mm and the 5 mm aerogel boards qualify, while the 3 mm board would allow the surface temperature to rise above 165 °C. Therefore, the 4 mm board is the thinnest available layer that satisfies the safety criterion with an acceptable margin.
I subsequently conducted a physical thermal-spread control experiment with a 4 mm aerogel board. The measured temperature curve was compared with the simulation and the earlier fitting result. The maximum temperature at the battery surface was about 142.1 °C, which is clearly below the dangerous threshold. The cell voltage displayed no evidence of internal short circuit. The 4 mm aerogel was thus confirmed as the optimal single-layer insulation structure for the given thermal abuse condition.
Heat Transfer and Flow Field Characteristics
The simulation was used to visualise physical quantities that are difficult to measure directly. The temperature distribution on the battery skin at 1200 s was highly uneven. The side facing the aerogel was the warmest, while the far side remained much cooler. Because of the low radial thermal conductivity of the jelly roll, the temperature difference between the front and rear surfaces exceeded 70 °C. Vertically, the battery surface also exhibited a small but distinguishable temperature gradient, caused by natural convection and the local heating position.
The boundary heat flux distribution showed that most of the heat is transferred through a band roughly equal to the diameter of the heating rod. The heat flux decreased rapidly with increasing distance from this central heating band. The silica aerogel not only reduced the total heat flux toward the battery but also significantly smoothed the temperature non-uniformity on the protected side.
Streamlines generated in the simulation illustrate the natural convection pattern inside the chamber. Hot air rises from the heating rod and aerogel surfaces. As the airflow reaches the top of the computational domain, it spreads laterally and descends along the side walls, forming a large recirculation loop. The air velocity near the heating rod was naturally higher than that near the battery because the elevated temperature near the rod produced stronger buoyancy. The maximum instantaneous velocity in the entire flow field was about 0.64 m/s. At a horizontal mid-plane of the battery, the maximum air speed near the heating rod reached roughly 0.4 m/s after 200 s, whereas the maximum air speed near the battery side was only about 0.14 m/s after the same period. The latter value corresponds to approximately 35% of the former, confirming that the aerogel barrier substantially suppresses both radiative and convective heat transfer toward the downstream cell.
These flow-field results also show why the boundary-layer mesh and transient solution were necessary. Natural convection has a measurable influence on the vertical temperature distribution of the battery surface. Although radiation is the dominant heat transfer mechanism at the high heater temperature, the contribution of convection cannot be ignored in an accurate predictive model.
Extension to a Battery Module
Module Configuration
To assess the practical effectiveness of the single-layer aerogel barrier inside an EV battery pack, I built a simulation model of a 3 × 3 module composed of nine 18650 cells. Four-millimetre aerogel boards were inserted between adjacent cells. The heating rod was placed at three different positions to emulate different failure locations within the module. In Case A, the heating rod was located at the centre of the nine-cell array. In Case B, it was located at one corner of the array. In Case C, it was placed on one side of the module. The battery temperatures in each case were monitored at selected cells.
The simulation results are summarised in Table 9.
| Heating position | Monitored cell | Final max temperature / °C |
|---|---|---|
| Case A: centre | Cell 1 | 60.18 |
| Cell 2 | 133.55 | |
| Case B: corner | Cell 1 | 28.30 |
| Cell 4 | 30.65 | |
| Cell 5 | 59.98 | |
| Cell 7 | 35.17 | |
| Cell 8 | 133.73 | |
| Case C: side | Cell 1 | 134.31 |
| Cell 3 | 60.08 | |
| Cell 4 | 132.65 | |
| Cell 6 | 30.69 | |
| Cell 7 | 34.69 |
Throughout all three heating arrangements, the highest cell surface temperature in the entire module remained below 134.31 °C. This value is safely below the internal short-circuit temperature observed in the unprotected single-cell tests. The cells that were separated from the heating rod by more than one aerogel layer remained at a temperature close to ambient, typically around 30 °C. In practical terms, this means that even if one isolated cell in a module fails, the single-layer aerogel barrier surrounding every individual cell is sufficient to prevent catastrophic propagation throughout the entire EV battery pack. The simulations thus provide strong evidence that a 4 mm ceramic-fibre silica-aerogel board can serve as an effective passive fire barrier in the module architecture of an EV battery pack.
Temperature and Flow Visualisation in the Module
The cross-sectional temperature field in the module confirms that the aerogel boards block a large fraction of the heat flux. The temperature drop across each 4 mm board is significant, and the heated side of the board can be several hundred degrees hotter than the protected side. In the horizontal plane, the temperature field is roughly symmetric about the heating source, but the vertical direction shows an upward skew due to natural convection. Air heated by the rod flows upward through the narrow channels between the cells and insulation layers. Near the top of the module, the air temperature and flow velocity are both higher than near the bottom. Nevertheless, the temperature rise at the protected cells remains modest, indicating that the convective contribution is effectively damped by the insulation layers and the narrow geometry of the module.
The streamlines around the module show that the overall natural convection recirculation inside the chamber influences the thermal environment of the cells. In the absence of forced convection, the primary cooling path for the far-field boundaries of the module is also natural convection. In a real EV battery pack, additional boundary cooling may be provided by a liquid cooling plate or jacket. The aerogel barrier then works in combination with the active cooling system to limit the consequences of a local thermal failure.
Three-Layer Insulation Structure
Design Idea
Although the 4 mm aerogel layer already offers satisfying protection, I wished to further improve the performance without adding too much thickness. To this end, I designed a three-layer insulation structure in which the aerogel board is sandwiched between two thin metal sheets. The metal sheets act as radiation shields, reflecting part of the incident radiative heat back toward the heat source. They also provide a more uniform temperature boundary on both sides of the aerogel, which improves the repeatability of the thermal protection. In this structure, the central aerogel layer retains its role as the principal thermal barrier, while the metal layers reduce the radiative load at the outer surfaces.
Experimental Study of the Metal-Layer Thickness
In the first series of tests, I used copper sheets of thicknesses 0.1 mm, 0.2 mm, 0.3 mm, and 0.4 mm. The insulation assembly was 90 mm × 90 mm in area. Thermocouples were mounted on the rear surface of the insulation board to measure the temperature response. A control case without any metal layers was also tested using the same 4 mm aerogel board. The heating time was again 1200 s.
The surface temperature of the heating rod reached about 630 °C after an initial transient and then stabilised. The back-surface temperature of the insulation assembly in the control case rose to about 309 °C after reaching a quasi-steady state. The three-layer assemblies reduced this rear-surface temperature substantially. Copper sheets of 0.2 mm, 0.3 mm, and 0.4 mm thickness produced very similar results. In the quasi-steady region, the differences among these three cases were only about 3 °C. The 0.1 mm copper sheets performed slightly worse, with a mean rear-surface temperature about 10 °C higher than that of the 0.2 mm case during the 800–1200 s period.
These results indicate that once the metal layer is sufficiently thick to be continuous and resistant to oxidation, additional thickness brings only marginal benefit. The thinner copper layer is more prone to oxidation and warp during heating, which increases its surface emissivity and reduces its ability to reflect infrared radiation. Thus, I chose 0.2 mm as the preferred metal-layer thickness for subsequent experiments, as it provides the required reflective performance with the smallest added mass.
Effect of Metal Material
In the second series of experiments, I compared copper (Cu), tungsten (W), iron (Fe), and Inconel 718 alloy sheets at a constant thickness of 0.2 mm. Each insulation assembly was first tested in a back-surface temperature-rise experiment. The copper and tungsten assemblies produced very similar rear-surface temperatures, with final average values around 151 °C near the end of the test. Iron gave a slightly higher rear-surface temperature of approximately 162.6 °C. Inconel 718 reached an average of 154.4 °C, but its temperature curve displayed a faster initial rise and reached the quasi-steady state earlier than the other materials. Except for tungsten, which required a noticeably longer time to reach thermal equilibrium, all materials exhibited similar qualitative behaviour.
The differences arise from the competing effects of surface emissivity, thermal diffusivity, and oxidation behaviour at elevated temperatures. Copper has a low emissivity and high thermal diffusivity, which allows the heat absorbed at the outer surface to spread quickly and re-emit from the entire area. Tungsten also has a low emissivity, but its response is slower. Iron oxidises more easily at high temperatures, causing its surface emissivity to increase over time. Inconel 718 has good oxidation resistance, but its surface emissivity is higher than that of polished copper in the infrared range. The measured data are summarised in Table 10.
| Metal material | Mean rise rate 200–400 s / °C s⁻¹ | Mean rear-surface temperature 1100–1200 s / °C | Time to steady state / s |
|---|---|---|---|
| Copper (Cu) | 0.251 | 151.41 | 660 |
| Tungsten (W) | 0.252 | 151.64 | 881 |
| Iron (Fe) | 0.290 | 162.56 | 707 |
| Inconel 718 | 0.335 | 154.35 | 564 |
In the subsequent thermal-spread control experiments with battery cells, copper again produced the lowest battery surface temperature. At the end of 1200 s, the battery surface temperature reached only about 91.6 °C when the three-layer assembly used 0.2 mm copper sheets. The tungsten assembly resulted in 98.3 °C, iron in 101.2 °C, and Inconel 718 in 100.5 °C. Among the four tested materials, copper is thus the best choice for the reflective metal layers. In addition to its low emissivity, copper is widely available and relatively easy to bond to the aerogel board. The only notable disadvantage is its higher density relative to some other candidate materials, which increases the total mass of the insulation structure. However, for the tested board area of 90 mm × 90 mm, the added mass of two 0.2 mm copper sheets was about 29 g, which remains acceptable in most EV battery pack designs.
Performance Comparison with Single-Layer Aerogel
The thermal-spread control experiments were compared between the single-layer 4 mm aerogel and the optimised three-layer structure with 0.2 mm copper sheets. The three-layer structure reduced the final battery surface temperature by 39.18 °C compared with the single-layer aerogel. This corresponds to a reduction of approximately 29.96% from the single-layer value. In addition, the three-layer structure eliminated the slight voltage drop observed in the single-layer experiments, which indicates a more stable cell operating state throughout the abusive heating test. The measured temperature and voltage histories show that the three-layer board not only reduces the peak temperature but also lowers the temperature-rise rate over the entire transient. Figure 5‑type comparisons between the single-layer and three-layer experiments were obtained with both the temperature curves and the average rise-rate curves. The three-layer insulation structure clearly outperforms the single-layer board of the same aerogel core thickness.
Mechanism of the Three-Layer Insulation Structure
The improved thermal-spread performance of the three-layer structure can be attributed to three mechanisms. First, the silica aerogel core itself has an extremely low thermal conductivity because of its nanoporous network structure. The ceramic fibres embedded in the aerogel improve its mechanical strength and high-temperature stability, but the dominant thermal resistance still arises from the solid backbone and the confined gas phase within the pores. As the temperature increases, radiation through the porous aerogel becomes more significant, which explains why the effective thermal conductivity of aerogel composites increases at elevated temperatures.
Second, the thin metal sheets act as radiation shields. The polished copper surface has a very low emissivity, typically on the order of 0.03 before oxidation. According to Kirchhoff’s law, a low-emissivity surface also has a low absorptivity for incident infrared radiation. Therefore, a large fraction of the radiative heat emitted by the heating rod is reflected away from the battery instead of being absorbed by the outer layer of the insulation board. The reduction in absorbed radiative flux leads directly to a lower outer-surface temperature of the aerogel and, in turn, a lower heat flux to the protected cell.
Third, the metal sheets reduce convective heat transfer near the heated surface. The metal layer smooths the local temperature distribution and reduces the air-temperature gradient that drives buoyant convection. In the simulation of the single-layer board, the aerogel already reduced the maximum velocity of the air on the battery side to about one-third of that on the heater side. The three-layer structure further suppresses this flow because the outer metal sheet reaches a more uniform temperature than the bare aerogel board. A smaller temperature difference between the metal surface and the surrounding air means a weaker buoyancy force and thus lower natural-convection heat transfer.
In summary, the three mechanisms work together: the aerogel core suppresses conduction, the metal sheets suppress radiation, and the combined structure suppresses natural convection. The result is a highly effective passive thermal barrier that can be installed inside an EV battery pack with only a small increase in overall thickness.
Conclusions and Outlook
Through a combination of experiments and simulations, I have developed a systematic understanding of thermal runaway in 18650 NCA lithium-ion cells and of passive thermal-spread control strategies suitable for vehicle battery modules. The main conclusions of my work can be summarised as follows.
First, thermal abuse caused by a high-temperature heating rod drives the lithium-ion cell sequentially through four stages: sustained heating, internal short circuit, safety-valve opening, and thermal runaway. In the fully charged state, the maximum surface temperature reached 786.41 °C and the maximum temperature-rise rate reached 167.88 °C/s. The short time interval between internal short-circuit onset and thermal runaway demonstrates the extreme violence of the reaction chain.
Second, a ceramic-fibre silica-aerogel board is an effective passive thermal barrier for controlling thermal spread. A 5 mm thick board successfully protected a battery for 1200 s of high-power heating, whereas a 1.5 mm board merely delayed thermal runaway by 357 s. The difference in performance is primarily due to the nonlinear effect of thickness on the total thermal resistance.
Third, numerical simulation allowed me to optimise the single-layer aerogel thickness without manufacturing every possible sample thickness. Using a cubic fitting model, I identified the 4 mm board as the thinnest available layer that satisfies a conservative safety criterion. The simulation results also provided valuable information about the temperature field, heat flux, and natural convection patterns that cannot be easily measured in experiments. Extending the simulation to a 3 × 3 module showed that the 4 mm board is capable of keeping all cells in the module below a hazardous temperature, even when the heating source is located in different positions. This supports the feasibility of using the same aerogel barrier architecture in a full-size EV battery pack.
Fourth, the three-layer insulation structure with 0.2 mm copper sheets on both sides of the 4 mm aerogel core offers the highest thermal-spread suppression performance among all tested configurations. It lowers the battery surface temperature by approximately 30% compared with the single-layer aerogel under the same thermal abuse condition. The combination of low-emissivity metal layers and a low-thermal-conductivity aerogel core provides a compact, passive, and reliable solution to the problem of thermal-runaway propagation.
Future work should extend these investigations to larger battery modules and full battery packs. In particular, cyclic thermal loading and mechanical vibration typical of vehicle operation should be considered to evaluate the long-term durability of the insulation structure. In addition, the aerogel barrier could be combined with an active cooling system to provide even greater protection under sustained thermal abuse. Another promising direction is to develop lighter sandwich structures that maintain the same protective performance with reduced overall mass, which would improve the gravimetric energy density of the EV battery pack.
In conclusion, the passive control of thermal spread in an EV battery pack is achievable using currently available composite materials. My study confirms that a well-designed aerogel-based insulation layer can dramatically reduce the thermal consequence of a single-cell thermal runaway event. The use of optimised single-layer or three-layer structures represents a simple yet robust strategy for improving the safety of electric vehicles and for protecting occupants and property from battery-induced fires.