Polymer-BN Composites as Thermal Interface Materials for Lithium-Ion Battery Modules: Experimental and Simulation Insights

Abstract

Efficient thermal management is critical for the safety and performance of lithium-ion battery (LIB) systems, particularly under high C-rate charge–discharge cycling. Here, we investigate two classes of polymer composite thermal interface materials (TIMs): graphene-PLA (GPLA) fabricated via 3D printing and boron nitride nanoplatelets (BN)-loaded thermoplastic polyurethane (TPU) composites with 20 and 40 wt.% BN content. To understand cooling dynamics, we developed a simple analytical model based on Newtonian heat conduction, predicting an inverse relationship between the cooling rate and the TIM thermal diffusivity. We validated this model experimentally using a six-cell LIB module equipped with active liquid cooling, and complemented it with finite-element simulations in COMSOL Multiphysics incorporating experimentally derived parameters. Across all approaches, analytical, numerical, and experimental, we observed excellent agreement in predicting the temperature decay profiles and inter-cell temperature differentials ($\mathrm{\Delta }T$). Charge–discharge cycling studies at varying C-rates demonstrated that high-diffusivity TIMs enable faster cooling but require careful design to minimize lateral thermal gradients. Our results establish that an ideal TIM must simultaneously support rapid vertical heat sinking and effective lateral thermal diffusion to ensure thermal uniformity. Among the studied materials, the 40% BN–60% TPU composite achieved the best overall performance, highlighting the potential of BN filler-engineered polymer composites for scalable thermal management in next-generation battery systems.

1. Introduction

As lithium-ion batteries (LIBs) become increasingly indispensable in electric vehicles (EVs), portable electronics, and stationary energy storage systems, their thermal management has emerged as a central engineering challenge [1,2,3,4,5,6,7,8]. Battery operation generates substantial heat, particularly under fast-charging or high-discharge conditions, leading to temperature gradients, accelerated degradation, and, in severe cases, thermal runaway [9,10]. Even a small (≤1 °C) but persistent gradient between cells can induce cell-to-cell capacity drift and uneven state-of-charge evolution ultimately affecting pack balancing. Thus, maintaining thermal uniformity and dissipating heat efficiently are critical for battery performance, safety, and lifespan.

In typical battery modules, heat is dissipated through passive and active cooling strategies, including airflow, liquid cooling manifolds, and heat pipes [2,3,4,5,11,12,13]. However, a significant thermal bottleneck persists at the interface between battery cells and their surrounding cooling infrastructure. Imperfect contact, surface roughness, and the low thermal conductivity of adhesives or structural materials contribute to significant thermal contact resistance [9,11,14,15,16,17,18]. This makes developing effective thermal interface materials (TIMs) crucial to advancing battery thermal management systems (BTMS).

TIMs are designed to bridge the thermal resistance gap between solid interfaces, such as battery casings and heat sinks. Polymer-based TIMs are often favored due to their mechanical flexibility, electrical insulation, and ease of processing [9,11,15,16,18,19]. Yet, the low intrinsic cross-plane thermal conductivity of most polymers (typically < 0.5 W m⁻¹ K⁻¹) severely limits their ability to conduct heat. To overcome this, high-conductivity fillers such as graphite, metal oxides, and carbon nanostructures like graphene are incorporated into the polymer matrix to form thermally conductive composites [20,21,22].

Graphene, with its exceptional in-plane thermal conductivity (up to 5000 W m⁻¹ K⁻¹), high aspect ratio, and tunable surface chemistry, has emerged as one of the most promising TIM fillers [18,23,24,25,26,27]. Its incorporation into polymer matrices can dramatically enhance thermal performance, particularly when oriented or networked to form continuous heat conduction paths [15,28,29,30,31,32,33]. However, graphene composites are also electrically conducting, which is a disadvantage in preventing undesired short circuits within the battery module.

Hexagonal boron nitride (h-BN), often dubbed “white graphene,” offers similarly high thermal conductivity along with electrical insulation, making it attractive for TIM applications where dielectric performance is important [17]. BN-polymer composites have demonstrated thermal conductivities up to 10 W m⁻¹ K⁻¹ depending on filler loading, morphology, and processing method [16,18,34,35,36,37]. A table summarizing prior results of thermal conductivity for various TIMs is presented in Table S1 [38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55]. Recent advances in filler surface functionalization, hybrid filler strategies, and anisotropic alignment have further improved the thermal transport efficiency of such composites.

While the thermal performance of individual materials is important, their integration into real-world systems poses additional design constraints. For example, TIMs must conform to irregular geometries, maintain thermal performance under cyclic loading, and avoid adding excessive mass or volume to the battery pack [11]. A comprehensive evaluation of TIMs requires multiscale assessment from intrinsic thermal properties such as thermal diffusivity to system-level metrics like inter-cell temperature gradients and transient cooling behavior [9]. Finite-element simulations, particularly those implemented in platforms such as COMSOL Multiphysics 6.2, have become essential in linking material properties with thermal field evolution in complex battery assemblies [35,56].

In this study, we investigated the fabrication and thermal performance of two classes of polymer composite TIMs: graphene–PLA (GPLA) fabricated by 3D printing, and boron nitride (BN)–loaded thermoplastic polyurethane (TPU) composites containing 20 and 40 wt.% BN. Building on prior reports of enhanced thermal conductivity in such systems [11,15,17,18,28,35,56], we conducted both experiments and simulations using two complementary setups designed to evaluate TIM performance under distinct thermal conditions. In the first configuration, individual cylindrical cells were coupled to an external heat source to quantify heat dissipation through each TIM. In the second, the TIMs were implemented in a compact $3\mathrm{s}2\mathrm{p}$ battery module, where the cells were charged and discharged at controlled C-rates. All experiments were performed with an active liquid-cooling manifold and external forced convection. For both setups, finite-element simulations were performed in COMSOL under identical boundary conditions to the experiments.

The electro-thermal model in COMSOL employed the standard LiMn₂O₄–graphite chemistry available in the built-in Lithium-Ion Battery Module to represent heat generation, while the electrochemical kinetics were not modeled explicitly. Instead, the heat source term was normalized for the capacity and C-rate of the NCA-based experimental cells, ensuring consistency of the total Joule-heating magnitude. The focus of the simulations was therefore comparative: to evaluate how temperature evolves with different TIMs rather than to reproduce charge–discharge voltage behavior. The cell and TIM material properties were assigned from experimentally measured or literature-reported values, and the analysis centered on the relative change in temperature for low- and high-diffusivity TIMs (PLA and 40 wt.% BN-TPU, respectively). The simulations closely reproduced the experimental temperature evolution, yielding excellent correlation for both the relative temperature difference $\mathrm{\Delta }T$ between TIMs and the cooling time constant $\tau$ extracted from the decay curves. The strong agreement between experiment and simulation confirms that the adopted comparative, heat-transfer-focused framework reliably captures the influence of TIM thermal diffusivity on module-level temperature uniformity.

While the comparative analysis highlights the consistency between experiment and simulation, it also reveals an intrinsic design trade-off in thermal interface engineering. A material with lower thermal diffusivity acts as a transient heat reservoir, storing heat and moderating short-term temperature spikes but offering slower lateral equilibration. Conversely, a high-diffusivity TIM rapidly spreads and dissipates heat, minimizing local hot spots yet transferring energy more quickly to the cooling infrastructure. In the present study, the GPLA system exemplifies the heat-storage regime, whereas the 40 wt.% BN–TPU composite represents the heat-dissipation regime with superior thermal uniformity across cells. This balance between transient thermal buffering and efficient dissipation defines the optimal operating window for polymer–composite TIMs in battery enclosures.

2. Materials and Methods

2.1. Materials

Polylactic acid or PLA filament was purchased from Hacthbox (Rowland Heights, CA, USA) while graphene-loaded PLA or GPLA filament was purchased from Blackmagic3D.com (Ronkonkoma, NY, USA). Thermoplastic polyurethane or TPU filament and TPU pellets were purchased from TCPoly (Pineville, NC, USA). Hexagonal boron nitride (h-BN) micropowder and N, N-dimethylformamide (DMF) were purchased from Sigma Aldrich (St. Louis, MO, USA).

2.2. Fabrication of Thermal Interface Materials

For preparing thermoplastic polyurethane or TPU-BN composites, TPU pellets were dried in a vacuum oven at 100 °C for 4 h. Exfoliated micro h-BN micropowder was dispersed in N, N-dimethylformamide (DMF) using a tip sonicator at 10 watts for 30 min. After sonication, the mixture was placed on a hot plate at 80 °C and magnetically stirred at 150 rpm. Dried TPU pellets were added to the mixture slowly. Subsequently, the mixture was stirred at 80 °C and 150 rpm for 4 h. Finally, the viscous mixture was cast onto a thin kitchen aluminum foil using an MTI doctor blade casting system to maintain a uniform thickness of 1.4 mm. The films were dried overnight at room temperature inside a chemical hood (120 cfm airflow) to remove the solvent slowly. Once dried, the TIMs were peeled from the foil and cut into 70 × 30 × 1.4 mm slabs.

We prepared multiple controls such as neat TPU, polylactic acid (PLA) and graphene-loaded PLA (GPLA) through 3D printing to validate trends against BN–TPU composites. The GPLA composites included as a reference material to benchmark the effect of a thermally conductive and electrically active filler against the insulating BN–polymer systems. Although not intended for direct battery use, GPLA provides a useful control for isolating filler-type effects on heat transport and interfacial coupling. For 3D printing, a 70 × 30 × 1.4 mm slab was designed using the geometry module of COMSOL Multiphysics 6.2. Upon finalizing the geometry, the design was exported and sliced in PrusaSlicer 2.9.4 software to generate the code for 3D printing. The exported code was then printed using a Prusa i3 MKS3+ 3D printer (Prusa by Josef Prusa, Prague, Czech Republic).

2.3. Characterization Techniques

All TIMs were characterized using thermogravimetric (TGA) analysis (TA instrument SDT Q600, New Castle, DE, USA) under nitrogen gas flow at 100 mL/min from room temperature to 600 °C with a ramp of 20 °C/min (Figure 1). Cross-plane thermal diffusivity was measured using the laser flash technique (Linseis LZT meter, Selb, Germany). Before diffusivity measurements, all samples were coated with graphite 33 spray to ensure uniform contact for heat flow.

To evaluate the performance of the TIMs, two complementary experimental setups were implemented using identical cooling and measurement configurations. In the first setup, six 18650-type cells (Panasonic NCR18650GA, 3300 mAh) were mounted in a custom aluminum manifold with internal coolant tubes (Figure 2a) similar in design to the liquid-cooling plates employed in commercial EV battery packs (e.g., Tesla Model 3). The liquid coolant (deionized water) was circulated through the manifold at a flow rate of 40 mL min⁻¹ using an electric pump. As shown in Figure 2, TIMs were placed between the cooling tubes and the battery surfaces on both sides of the module to ensure uniform thermal contact. Six K-type thermocouples were inserted to monitor the temperature of each cell, and data were recorded using the PicoLog 6 software. In addition, an infrared camera (FLIR E60, Wilsonville, OR, USA) was used to capture thermal images at different stages of the heating and cooling process. The battery module was heated externally using a hot plate to approximately 45 °C and subsequently cooled inside a hood with controlled airflow at 120 cfm. All measurements were performed in triplicate ($n=3$) to ensure statistical reproducibility.

In the second setup, the TIMs were tested under active electro-thermal cycling in a prototype battery pack configured as $3\mathrm{s}2\mathrm{p}$ using six Panasonic NCR18650GA cells. The pack was enclosed in an environmental chamber (Neware) and fitted with the same aluminum coolant manifold, thermocouple placement, and coolant flow conditions as in the external-heating setup. The pack was charged and discharged continuously at different C-rates using a Neware CE6000 (20 A/60 V) testing system (Shenzhen, China) while monitoring the surface temperatures of each cell. The water coolant was circulated continuously during operation (40 mL min⁻¹), and the chamber was sealed to maintain consistent ambient conditions. Tests were repeated for both TIM configurations (PLA and 40 wt.% BN–60 wt.% TPU) following the same experimental protocol. This dual approach enabled the evaluation of TIM performance under both externally induced thermal loading and realistic charge–discharge cycling conditions, providing a comprehensive assessment of thermal dissipation and uniformity.

Scanning electron microscope images, presented in Figures S1 and S2 of the supporting information, were obtained using a Hitachi 6600 SEM (Schaumburg, IL, USA). High resolution X-ray diffraction (HR-XRD) is performed using a RIGAKU Ultima IV diffractometer (Tokyo, Japan) employing Cu K$\alpha$ radiation, on powder samples that were held by a standard Al sample holder. Quantitative analysis using Rietveld refinement is performed on the XRD peaks using PDXL 2 software. Atomic force microscopy (AFM) measurements were performed in a non-contact mode using AIST-NT SPM Smart system (HORIBA, Kyoto, Japan) and cantilevers (HQ: NSC14/Al BS-50) from Micromasch. AIST-NT image analysis and processing (Version 3.2.14) software was used for AFM image analysis. Malvern Zetasizer 90 (Westborough, MA, USA) was used for dynamic light scattering measurements. Based on the DLS measurements, the average lateral size of BN particles is 745 ± 92 nm, which is consistent with the average size deduced from AFM measurements that showed an average lateral size of 600–800 nm with a thickness of 70–100 nm (see Figure S3 in the supporting information). Differential scanning calorimetry was performed to extract the specific heat values (Figure S4) using TA instruments Discovery DSC (New Castle, DE, USA) from 25 to 600 °C.

2.4. COMSOL Modeling

The computational modeling of heat-transfer behavior was carried out using COMSOL Multiphysics 6.2 employing the finite-element method (FEM). The goal of the simulations was not to model electrochemical reactions or voltage behavior, but rather to evaluate the thermal efficiency of different TIMs under controlled and comparable boundary conditions. Accordingly, the built-in electro-thermal Li-ion Battery Module (liion) was used with the standard LiMn₂O₄ (LMO)–graphite chemistry available in the COMSOL materials library [57,58,59]. This configuration provides a well-validated electro-thermal coupling framework for heat generation without explicitly solving the full electrochemistry. To represent the NCA-based Panasonic NCR18650GA cells used experimentally, the model was normalized to their nominal capacity and C-rate (3.7 V, 3300 mAh, 1C = 3.3 A) so that the total heat generation corresponded to the actual operating conditions of the commercial cells. Because all TIMs and boundary conditions were held identical across simulations, this normalization allows for a direct comparative assessment of thermal performance without affecting the relative temperature trends.

Two simulation configurations were developed, mirroring the experimental setups described earlier. In the first configuration, six cylindrical 18650 cells were modeled within an aluminum cooling manifold subjected to an external heat flux to replicate the controlled-heating experiments. The coolant flow and convective boundary conditions were identical to those in the laboratory tests. Additionally, individual-cell simulations were performed to study the effect of C-rate on temperature evolution by charging single cells independently at different C-rates. In the second configuration, a $3\mathrm{s}2\mathrm{p}$ battery module was modeled, with all six cells electrically connected in series–parallel and operated under a continuous charge–discharge protocol. This setup replicates the electro-thermal-cycling experiments performed in the environmental chamber.

The electro-thermal model was coupled to the Heat Transfer in Solids and Fluids (ht) module using the spatially averaged heat-generation rate $Q_{h\left(\mathrm{avg}\right)}$ and temperature $T_{\mathrm{avg}}$ from the liion module. The coolant flow through the aluminum manifold was implemented using the Laminar Flow (spf) module with a non-slip boundary condition and a volumetric flow rate of matching the experimental value of 40 mL min⁻¹. The heat-transfer and flow physics were coupled through the Non-Isothermal Flow (nitf) interface to capture the conjugate heat exchange between the cells, TIMs, and coolant. For the second configuration, battery pack (bp) interface was used instead, with electrochemical heating(ech) as the multiphysics to account for the.

The geometric dimensions and material properties of the TIMs (70 × 30 × 1.4 mm) matched those used experimentally. Anisotropic thermal conductivities were assigned to the cylindrical cells in a local coordinate system to approximate the spiral-wound electrode geometry. Normal meshing (maximum element size 11.4 mm) was applied to the solid domains, and finer meshing (maximum element size 5.3 mm) was used along the coolant tube surfaces to resolve temperature gradients accurately. The simulations yielded transient temperature profiles for all cells and TIM configurations, which were then analyzed to extract the cooling time constant $\tau$, the inter-cell temperature differential $\mathrm{\Delta }T$, and relative temperature differences with different TIMs.

3. Results

Figure 1a presents the TGA curves for five materials: neat PLA, GPLA, neat TPU, and TPU loaded with 20 wt.% and 40 wt.% BN. All samples were heated under nitrogen atmosphere up to 600 °C, and weight loss was monitored to evaluate thermal degradation behavior and filler content. Neat PLA and GPLA both exhibit a sharp one-step degradation between 300 °C and 400 °C, consistent with known depolymerization and chain scission mechanisms of PLA [11]. However, the GPLA sample demonstrates a slightly earlier onset of decomposition and retains significantly more mass at high temperature due to the presence of graphene fillers. Specifically, at 450 °C, PLA retains less than 2% of its initial mass, whereas GPLA retains approximately 15%. This difference indicates that there is ∼15 wt.% of graphene incorporated into GPLA. Neat TPU shows a broader and more gradual weight loss profile with a single step beginning near 300 °C, and a residual mass of around 45% at 450 °C. Such a single-step decomposition process is characteristic of the breakdown of both hard and soft segments in the polymer matrix. The addition of BN significantly alters this behavior. The 20%BN–80%TPU composite exhibits a two-step decomposition with a second decomposition onset at 400 °C and retains approximately 52% of its initial mass at 450 °C. The 40%BN–60%TPU sample demonstrates the highest thermal stability among all samples, retaining nearly 70% of its mass at the same temperature. The two-step decomposition is more evident in 40%BN–60%TPU sample. The progressive increase in residue with h-BN loading is indicative of the high thermal stability of h-BN, which acts as an inert filler resisting oxidative or thermal degradation [16]. The emergence of a second decomposition step in BN–TPU composites at 400 °C can be attributed to interfacial interactions between the BN fillers and the TPU chains. The high thermal stability and inertness of BN not only provide a thermal shielding effect but also restrict the mobility of TPU chains through non-covalent interactions such as van der Waals forces or hydrogen bonding (e.g., N–H⋯BN). These interfacial effects delay the decomposition of segments closely associated with BN surfaces, resulting in the observed higher-temperature degradation step. Therefore, the two-step decomposition is indicative of filler-mediated stabilization and heterogeneous breakdown of the polymer matrix in the presence of thermally robust BN domains.

Figure 1b shows the thermal diffusivity of the same five samples as a function of temperature up to 70 °C. The diffusivity data reveals a clear hierarchy in thermal transport performance, strongly dependent on the type and concentration of thermally conductive fillers. The highest thermal diffusivity is observed in the 40%BN–60%TPU composite, with value $2.32\times {10}^{-2}$ cm² s^–1 at a temperature of 30 °C. This value is more than an order of magnitude higher than that of the base TPU, which maintains a diffusivity near $3.5\times {10}^{-3}$ cm² s^–1. The 20%BN–80%TPU composite shows intermediate performance, with a diffusivity of approximately $8.5\times {10}^{-3}$ cm² s^–1. This behavior is consistent with general percolation theory for platelet-based composites, where continuous thermally conductive pathways typically emerge near one-third filler loading (30–35 wt.%), as reported for polymer systems with platelet shaped fillers such as h-BN and graphene [18,31]. Below this threshold (as in the 20%BN–80%TPU composite), filler particles are isolated and heat transfer remains matrix-dominated. Above the threshold (as in the 40%BN–60%TPU composite), an interconnected BN network enhances phonon transport and suppresses interfacial resistance. The GPLA sample exhibits only a marginal improvement over neat PLA, with diffusivity values of about $3.0\times {10}^{-3}$ cm² s^–1 compared to $2.1\times {10}^{-3}$ cm² s^–1 for PLA. This modest increase suggests graphene in PLA matrix (15 wt.% based on TGA in Figure 1a) is suboptimal for effective thermal transport. The enhancement of thermal diffusivity in TIMs after incorporation of BN filler into the neat polymer matrix is also observable in TIM IR images different temperatures (see supporting information Figure S6).

TIM expeimental characterization using the first setup with an external heat source: As mentioned in Section 2, two complementary experimental setups were implemented using identical cooling and measurement configurations. Below, we discuss the results from the first setup where a module comprising six cylindrical Li-ion cells was assembled with the TIM inserted between the cells and the liquid cooling manifold (Figure 2a,b). The module was subjected to controlled external heating inside a forced-air convection hood with an airflow rate of 120 cfm. Heating was continued until the average temperature of the battery module reached approximately 45 °C, at which point the heat source was turned off, initiating the cooling phase. Each cell was attached with an independent thermocouple to monitor transient temperature profiles throughout the heating and cooling cycles. The aqueous liquid coolant was pumped through the Al manifold throughout the heating and cooling cycles. Typical IR images of the cells at different stages of heating are shown in Figure 2c,d.

Figure 3a shows the representative thermal response for all six cells when PLA was used as the TIM. Although all cells were exposed to the same environmental conditions, the temperature evolution was non-uniform, with distinct differences in heating rates and peak temperatures. Notably, Battery-1 reached the highest temperature while Battery-6 remained significantly cooler, exhibiting a temperature difference of nearly 4 °C at the peak. These differences reflect the anisotropic thermal conductance within the module and asymmetries in cell-to-manifold contact. Figure 3b shows the average temperature ($\overline{T}$) of the entire module, obtained from the temperature readings from all six thermocouples. The average curve shows a smooth and symmetric profile with a clear demarcation between the heating and cooling phases. To evaluate thermal uniformity within the module, the temperature difference $\mathrm{\Delta }{T}_i\left(t\right)=|T_i\left(t\right)-\overline{T}\left(t\right)|$ was computed for each cell i with respect to the instantaneous average temperature $\overline{T}\left(t\right)$. The results are plotted in Figure 3c. The $\mathrm{\Delta }T$ curves show that thermal nonuniformity peaks near the end of the heating phase, where some cells lag in temperature rise compared to others. During the cooling phase, $\mathrm{\Delta }T$ gradually decreases as heat is more uniformly extracted via the manifold, but notable differences persist over the full transient period.

The transient cooling behavior of the battery module, following the removal of the external heat source, was analyzed using the following thermal conduction model. TIM serves as the dominant path for heat dissipation from the battery cells to the cooling manifold. Assuming negligible heat generation during the cooling phase and lumped thermal capacity of the cells, the rate of temperature decay can be modeled by equating the rate of internal energy loss to the conductive heat flux through the TIM. The rate of heat loss from the battery can be expressed as

$$-\dot{Q}=C_{p,\mathrm{dev}}{M}_{\mathrm{dev}}{\displaystyle \frac{dT}{dt}}$$

(1)

where $C_{p,\mathrm{dev}}$ is the specific heat capacity and $M_{\mathrm{dev}}$ is the effective thermal mass of the battery module. In the absence of a cooling liquid, the heat flux through the TIM is modeled as

$$\dot{Q}={\displaystyle \frac{\kappa A_{\mathrm{TIM}}}{d_{\mathrm{TIM}}}}(T-T_{\infty })$$

(2)

where $\kappa$ is the thermal conductivity of the TIM, $A_{\mathrm{TIM}}$ is the contact area, $d_{\mathrm{TIM}}$ is the TIM thickness, T is the instantaneous temperature of the battery, and $T_{\infty }$ is the steady-state temperature.

Equating these expressions yields the governing differential equation for temperature decay:

$$C_{p,\mathrm{dev}}{M}_{\mathrm{dev}}{\displaystyle \frac{dT}{dt}}=-{\displaystyle \frac{\kappa A_{\mathrm{TIM}}}{d_{\mathrm{TIM}}}}(T-T_{\infty })$$

(3)

which simplifies to

$${\displaystyle \frac{dT}{dt}}=-{\displaystyle \frac{\kappa A_{\mathrm{TIM}}}{C_{p,\mathrm{dev}}{M}_{\mathrm{dev}}{d}_{\mathrm{TIM}}}}(T-T_{\infty })$$

(4)

Introducing the thermal diffusivity $\alpha =\kappa /\left(\rho C_{p,TIM}\right)$, the time constant $\tau$ governing exponential decay is given by

$$\tau ={\displaystyle \frac{C_{p,\mathrm{dev}}{M}_{\mathrm{dev}}{d}_{\mathrm{TIM}}}{{\rho }_{\mathrm{TIM}}{C}_{p,\mathrm{TIM}}{\alpha }_{\mathrm{TIM}}{A}_{\mathrm{TIM}}}}$$

(5)

showing that $\tau$ is inversely proportional to the thermal diffusivity ${\alpha }_{\mathrm{TIM}}$ of the interface material. Thus, the solution to the temperature decay equation is an exponential approach to $T_{\infty }$:

$$T\left(t\right)=T_{\infty }+(T_0-T_{\infty })e^{-t/\tau }$$

(6)

where $T_0$ is the initial temperature at the start of the cooling phase. Equation (6) was used to fit the experimental cooling data for each TIM configuration. The extracted decay constant $\tau$ provides a quantitative measure of the rate at which the battery module sheds heat. All the fit values are provided in Table S2 in the supporting information. Lower values of $\tau$ correspond to faster cooling and more thermally effective TIMs. The fitted results were subsequently compared across different TIM materials and correlated with their independently measured thermal diffusivities to verify the predicted inverse relationship $\tau \propto 1/{\alpha }_{\mathrm{TIM}}$. A representative plot (see Figure 3d) confirms this trend.

It is important to note that the analytical model presented above isolates the dominant contribution of heat conduction through the TIM, while assuming other paths and contact resistances to be constant. In the actual experimental configuration, heat is dissipated simultaneously through (i) the liquid coolant flowing in the Al manifold and (ii) external forced convection at an airflow rate of 120 cfm. The complete heat-transfer network therefore consists of the TIM conduction resistance in series with contact, aluminum, and coolant-side resistances, acting in parallel with the external convective branch. To account for these effects, Equation (3) can be generalized as

$${\dot{Q}}_{\mathrm{tot}}=(T-T_{\infty })\left[{\displaystyle \frac{1}{{\displaystyle \frac{d_{\mathrm{TIM}}}{k_{\mathrm{TIM}}{A}_{\mathrm{TIM}}}}+R_{c,\mathrm{cell}|\mathrm{TIM}}+R_{c,\mathrm{TIM}|\mathrm{manifold}}+R_{\mathrm{Al}}+{\displaystyle \frac{1}{h_{\mathrm{liq}}{A}_{\mathrm{liq}}}}}}+h_{\mathrm{ext}}{A}_{\mathrm{ext}}\right],$$

(7)

where $R_{c,\mathrm{cell}|\mathrm{TIM}}$ and $R_{c,\mathrm{TIM}|\mathrm{manifold}}$ denote the contact resistances at the respective interfaces, $R_{\mathrm{Al}}$ is the conduction resistance of the Al manifold, and $h_{\mathrm{liq}}{A}_{\mathrm{liq}}$ and $h_{\mathrm{ext}}{A}_{\mathrm{ext}}$ represent the convective heat-transfer coefficients and effective areas for the liquid and air cooling paths, respectively. The corresponding cooling time constant can then be expressed as

$$\tau ={\displaystyle \frac{C_{p,\mathrm{dev}}{M}_{\mathrm{dev}}}{G_{\Vert }}}={\displaystyle \frac{C_{p,\mathrm{dev}}{M}_{\mathrm{dev}}}{{\displaystyle \frac{1}{R_{\mathrm{series}}}}+h_{\mathrm{ext}}{A}_{\mathrm{ext}}}},$$

(8)

where $R_{\mathrm{series}}={\displaystyle \frac{d_{\mathrm{TIM}}}{k_{\mathrm{TIM}}{A}_{\mathrm{TIM}}}}+R_{c,\mathrm{cell}|\mathrm{TIM}}+R_{c,\mathrm{TIM}|\mathrm{manifold}}+R_{\mathrm{Al}}+{\displaystyle \frac{1}{h_{\mathrm{liq}}{A}_{\mathrm{liq}}}}$ and $G_{\Vert }$ is the total parallel conductance. Under identical cooling conditions across all experiments, the comparative dependence of $\tau$ on $1/{\alpha }_{\mathrm{TIM}}$ remains valid, with the slope determined by the TIM properties and the intercept representing the cumulative parasitic resistances (see the inset in Figure 3d).

Figure 4 presents the time-resolved temperature deviations $\mathrm{\Delta }{T}_i\left(t\right)=|T_i\left(t\right)-\overline{T}\left(t\right)|$ for each of the six cells in the attery module using five different TIMs: (a) PLA, (b) GPLA, (c) TPU, (d) 20%BN–80%TPU, and (e) 40%BN–60%TPU. The metric $\mathrm{\Delta }T$ quantifies the deviation of each cell’s temperature from the instantaneous module average $\overline{T}\left(t\right)$, providing a dynamic measure of inter-cell thermal uniformity. Interestingly, the PLA TIM—despite having the lowest thermal diffusivity among the tested materials—exhibits the narrowest spread in $\mathrm{\Delta }T$, with all six cells maintaining relatively close thermal trajectories. The peak $\mathrm{\Delta }T$ remains below 3 °C throughout the thermal cycle. This result appears counterintuitive given PLA’s poor thermal conductivity. However, it can be rationalized by considering the relatively uniform thermal impedance across all cell–manifold interfaces: in the absence of a highly conductive path, each cell dissipates heat at a similar, albeit slow, rate (as indicated by a high $\tau$ for PLA in Figure 3d). The system behaves almost adiabatically on short time scales, and thermal gradients develop primarily due to intrinsic differences in cell positioning or contact pressure rather than material-mediated heat spreading.

In contrast, the introduction of higher-diffusivity TIMs such as GPLA and TPU increases the magnitude and spread of $\mathrm{\Delta }T$. Notably, the GPLA TIM (Figure 4b) exhibits the largest inter-cell thermal heterogeneity, with Battery-6 showing a pronounced deviation exceeding 5 °C. This suggests that while the thermal conductivity of GPLA is higher than PLA, it remains insufficient to rapidly redistribute heat, and may accentuate local differences arising from subtle variations in thermal contact resistance. Similarly, the TPU-only TIM (Figure 4c) produces a broader $\mathrm{\Delta }T$ spread than PLA, albeit less pronounced than GPLA. These observations underscore the sensitivity of thermal equilibration to both filler dispersion and the quality of interface coupling between cells and the cooling manifold. For the 20%BN–80%TPU system (Figure 4d), peak $\mathrm{\Delta }T$ values slightly increased, with Battery-1 deviating significantly from the average. In the 40%BN–60%TPU composite (Figure 4e), the inter-cell temperature differences become markedly lower, and the spread among all six cells narrows considerably. This suggests that only beyond a critical threshold of thermal diffusivity does the TIM effectively serve its homogenizing function, enabling lateral heat spreading across the module to suppress local temperature peaks.

It should be noted that an ideal TIM should simultaneously exhibit both a high rate of cooling (see Figure 3d) and a low spread of $\mathrm{\Delta }T$, which is achieved in our case through 40%BN–60%TPU composites. The physical mechanism underlying these trends can be understood in terms of the competition between two thermal processes: (1) vertical heat sinking into the cooling manifold, and (2) lateral thermal diffusion across adjacent cell interfaces via the TIM. At low diffusivity (e.g., PLA), both mechanisms are weak, resulting in symmetric but slow heat loss across the module. As diffusivity increases without reaching the percolation threshold (e.g., GPLA, 20%BN–TPU), lateral diffusion becomes anisotropic and non-uniform, which can amplify local hotspots, especially near the geometric or thermal boundaries of the module. Once the TIM achieves sufficiently high diffusivity (e.g., 40%BN–TPU), lateral heat transfer dominates, allowing efficient redistribution of thermal energy from warmer to cooler regions and thus reducing $\mathrm{\Delta }T$. Edge effects are also evident in these data due to the finite size of the battery module and the small number of monitored cells (N = 6). Cells located at the periphery (e.g., Battery-1 or Battery-6) often show the largest deviation from the average temperature. This behavior is consistent with the expectation that edge cells experience asymmetric thermal environments—differing airflow exposure, partial insulation, or varying proximity to the cooling manifold boundaries. To further elucidate the relationship between TIM thermal properties and inter-cell thermal uniformity, we extracted the peak values of $\mathrm{\Delta }T$ for each configuration and plotted them against the measured thermal diffusivity of the corresponding materials, as shown in Figure 4f. Error bars represent the standard deviation of the peak $\mathrm{\Delta }T$ values across three replicates for each TIM. In summary, the performance of a TIM involves a fundamental trade-off between heat storage and heat dissipation. Materials with low thermal diffusivity can buffer transient heat and delay propagation, whereas those with higher diffusivity promote rapid heat spreading and minimize local temperature gradients. In the present study, GPLA represents a heat-storage-dominated regime, while BN–TPU (particularly 40 wt.%) exemplifies a high-diffusivity, heat-dissipating regime that enhances thermal uniformity across cells.

COMSOL simulation of TIM performance using the first setup with an external heat source: To evaluate the dynamic thermal response of the battery module under practical operating conditions, COMSOL Multiphysics simulations were conducted across a range of charge–discharge (CD) rates (1C, 2C, 4C, and 8C) [2,60,61,62,63,64,65]. A six-cell Li-ion battery module was subjected to 10 continuous charge–discharge cycles at varying C-rates (1C, 2C, 4C, and 8C), followed by a 2.5-h cooling period with various TIMs placed between the cells, as shown in Figure 5. Figure 5a presents a representative image of the temperature profile at the end of the first cycle using PLA as the TIM. The thermal distribution among the cells is relatively symmetric. However, we notice an increase in the cell temperature, as evidenced by different colors of the liquid-cooling manifold (hot pink) compared to the cells (golden yellow). Figure 5b–f depict the temperature profiles at the end of the tenth cycle for the different TIMs. For PLA (Figure 5b), a significant temperature increase is evident, with pronounced hotspots in TIM toward the left edge of the cell. The GPLA and TPU TIMs (Figure 5c,d) demonstrate improved thermal performance relative to PLA, but still exhibit notably asymmetry as noticed from different colors for the liquid-cooling manifold (purple) and the cells (orange for GPLA and hot pink for TPU). These results are consistent with their moderate thermal diffusivities, which enable some degree of vertical heat sinking but are insufficient for complete lateral thermal equilibration. The 20% BN–TPU composite (Figure 5e) provides a further reduction in maximum temperature and a modest improvement in uniformity. The most striking improvement is observed with the 40% BN–TPU composite (Figure 5f), where the temperature field is both lower in magnitude and more homogeneous with the liquid-cooling manifold and the cells both showing similar color.

To analyze more quantitatively, we present the average temperature profiles in Figure 6. Each subplot compares the temperature rise for different TIM configurations: PLA, GPLA, TPU, 20% BN–TPU, and 40% BN–TPU. Across all conditions, a clear and systematic increase in temperature with increasing C-rate is observed, consistent with the enhanced internal heat generation governed by the relation $Q\propto I^{2}R$, where I is the current and R is the internal resistance of the cell. At low C-rates (e.g., 1C in Figure 6a), the thermal load is moderate leading to maximum temperature ≤22 °C. At 1C, all TIMs appear reasonably effective in preventing excessive temperature rise, although subtle differences in performance are already discernible with 40% BN–TPU showing the best performance, as expected. As the C-rate increases (particularly at 4C and 8C) in Figure 6b–d, the distinctions among TIMs become significantly more pronounced. Notably, the PLA TIM, which exhibits the lowest thermal diffusivity among the tested materials, results in the highest average temperatures at every C-rate. In contrast, the 40% BN–TPU composite consistently demonstrates the lowest temperature profiles across all C-rates, highlighting the role of enhanced thermal diffusivity in facilitating vertical heat sinking and efficient thermal management. Furthermore, all materials exhibit a thermal staircase in peak temperature with increasing C-rate, indicating cumulative heating effects due to incomplete thermal recovery between successive cycles.

At the highest C-rate of 8C, PLA exhibited the largest $T_{avg}$, reaching slightly above 50 °C over the course of the cycles (Figure 7a). In contrast, the 40%BN–TPU TIM, which demonstrated the highest experimental thermal diffusivity, limited the thermal rise to below 50 °C, despite identical electrochemical and convective boundary conditions. Figure 7b–e present the average inter-cell temperature deviation, calculated as $\delta {\overline{T}}_{CD}$=$\mathrm{\Sigma }(T_i-T_{avg})/6;(i=1,..,6)$, for each TIM configuration across charge–discharge cycles at different C-rates. Across all C-rates, a clear hierarchy in $\delta {\overline{T}}_{CD}$ emerges. Low-diffusivity TIMs such as PLA and GPLA exhibit the lowest inter-cell deviations in contrast to high-diffusivity TIMs (similar to results discussed in Figure 4). As discussed earlier, this can be rationalized in terms of the balance between vertical heat transfer to the manifold vs. the lateral heat spreading. While high diffusivity TIMs enable faster heat decay through vertical heat transfer (cf. Figure 3d), there is asymmetric lateral heat diffusion resulting in higher inter-cell deviations.

TIM performance using the second setup with a 3s2p module: To gain a deeper understanding of the effect of TIMs on battery pack thermal management and compare with COMSOL simulation, we designed a 3s2p battery pack with an aluminum coolant tube and TIMs (PLA and 40% BN + 60% TPU) configuration in COMSOL Multiphysics. The simulation was run at different charge–discharge rates. The respective 3D temperature mappings are shown in Figure 8. A significant decline in the average surface temperature was observed when 40%BN+TPU was used as the TIM material (Figure 8c,d) compared to PLA (Figure 8b,e).

Figure 9a shows the COMSOL simulated results for the average temperature of the battery pack at 1.4 C-rate. Clearly, the average temperature decreases when using BN-filled TPU TIM compared to PLA. To compare with experimental results, we measured temperature of six 18650 Li-ion batteries configured in 3s2p (along with a coolant tube and TIMs) similar to the simulation design. Each battery in the pack was connected with a thermocouple to record the temperature. The experiment was run at a continuous constant current charge–discharge rate of 1.4 C for 10 cycles. In Figure 9b, the average temperature of the battery pack is plotted after 10 cycles with neat PLA and 40% BN + 60% TPU TIMs. The results show that the average battery pack temperature reaches approximately 43 °C with PLA. In contrast, with BN-filled TPU TIM, the average pack temperature is recorded at approximately 38 °C. The percent difference between 40% BN + 60% TPU and PLA, observed in both experiment and simulations, are shown in Figure 9c. The observed trend reveals a remarkably strong relationship between the experimental and simulated results with a Pearson correlation coefficient of 0.9775. This indicates excellent linear agreement in their trends, where 97.75% of the variance in one series is explained by the other, demonstrating that the sequences evolve in near-perfect lockstep. Furthermore, the normalized correlation (using z-scores to focus solely on scale-invariant shape and pattern) mirrors this value at 0.9775, underscoring the simulation’s fidelity in capturing the temporal dynamics of the experimental data [66,67].

Figure 10 compares the experimental and simulated thermal responses of the $3\mathrm{s}2\mathrm{p}$ battery module under continuous charge–discharge cycling at different C-rates. Panels (a)–(c) present the percent temperature difference between the high-diffusivity (40 wt.% BN–TPU) and low-diffusivity (PLA) TIMs at 1.35C, 2C, and 2.7C, respectively. In both experiment and simulation, the BN–TPU composite consistently exhibits lower average cell temperatures throughout the cycling process, with the difference becoming more pronounced at higher current loads. The time-dependent profiles show strong overlap between the simulated and measured percent differences, confirming that the COMSOL model accurately captures the transient thermal behavior of the module. As the C-rate increases, the average temperature of the cells rises due to enhanced Joule heating, whereas the relative temperature reduction provided by the BN–TPU TIM ($T_{\mathrm{BN}-\mathrm{TPU}}-T_{\mathrm{PLA}}$) remains consistently negative, indicating superior heat dissipation (Figure 10d,e). The excellent agreement between experiment and simulation, within ±10% for the relative temperature difference, demonstrates the validity of the capacity- and C-rate-normalized electro-thermal model. Together, these results confirm that the high-diffusivity BN–TPU composite provides more effective heat removal and temperature uniformity across the module at all tested operating conditions.

4. Conclusions

We systematically evaluated the thermal performance of graphene-PLA (GPLA) and boron nitride (BN)-loaded TPU composites as thermal interface materials (TIMs) for lithium-ion battery modules. Material characterization confirmed that the incorporation of high thermal conductivity fillers significantly enhanced both thermal stability and thermal diffusivity. Through a combination of analytical modeling, finite-element simulations in COMSOL Multiphysics, and experimental thermal cycling studies, we established a direct inverse relationship between the TIM thermal diffusivity and the heat decay rate of the battery module. Our analysis further revealed that optimal TIMs must balance vertical heat sinking into the cooling infrastructure with lateral thermal diffusion across adjacent cells to minimize inter-cell temperature gradients. Among the materials studied, the 40%BN–60%TPU composite demonstrated superior performance, achieving both rapid cooling and thermal uniformity. These findings underscore the potential of filler-engineered, additive-manufactured polymer composites for scalable, efficient thermal management in high-power lithium-ion battery applications.