1. Introduction

As AI computing performance continues to scale, chip dimensions and areal power density increase in tandem^[1-3]. State-of-the-art AI accelerators now exhibit thermal design powers approaching or exceeding 1 kW, with die areas expanding to ~10³ mm², resulting in markedly elevated heat flux densities^[4-6]. These extreme thermal loads and large form factors induce pronounced die-package warpage during thermal cycling, which often reaches tens to hundreds of micrometers, with the most severe deformation near die edges. Such deformation amplifies stress concentration and the risk of thermomechanical fatigue failure. Accordingly, chip-scale thermal interface materials (TIMs)^[7,8] are required to possess a sufficiently low modulus to accommodate large warpage-induced deformation and strong interfacial adhesion to suppress delamination under cyclic loading, thereby maintaining a stable thermal conduction pathway^[9-12].

Silicone-based^[13-20] thermal gels, widely used as dispensable gap fillers, offer good process compatibility, mechanical compliance, simple fabrication and low cost. However, their intrinsically low thermal conductivity results in an overall thermal resistance typically exceeding 0.05 K cm² W^-1, and prolonged exposure to high-temperature often leads to material hardening. Oriented carbon-based^[21-26] pads exhibit high intrinsic thermal conductivity and low thermal resistance, while tolerating compressive strains of ~40-50%. But they lack intrinsic interfacial adhesion and thus depend on external mechanical pressure. Under tensile conditions induced by edge warpage in large-area packages, interfacial debonding may occur and disrupt heat-transfer pathways. Their reliance on highly conductive graphite films also increases cost and fabrication complexity, precluding large-area dispensing achievable with silicone gels. Metal-based thermal interface materials, such as indium foils, provide high intrinsic thermal conductivity and a relatively low modulus, and can form strong interfacial bonds after reflow. However, the low melting point of indium (156 °C) results in repeated melting during each reflow cycle, readily generating interfacial voids and defects. Moreover, their implementation requires backside metallization with multiple deposited layers, substantially increasing process complexity. These three technological routes are being pursued in parallel, with oriented carbon pads and metal-based^[27-31] TIMs attracting growing attention in advanced packaging. Despite their superior intrinsic thermal conductivity, carbon- and metal-based TIMs exhibit poorer mechanical compliance than polymer-based counterparts. As AI chip die sizes and package dimensions continue to expand, edge warpage becomes more pronounced, creating a demand for TIMs that accommodate both compressive and tensile deformation while retaining viscoelasticity. In this context, the development of new polymer-based thermal interface materials is critical for effective thermal management in large-area AI chips.

Polymer-based TIMs^[32-37] are dominated by silicone formulations due to their high flexibility and chemical stability. Despite decades of optimization, the performance remains intrinsically constrained. Specifically, silicone backbones exhibit weak interfacial interactions with common substrates. To improve their adhesion, promoters are typically required, but that will inevitably increase the elastic modulus, and preclude the simultaneous realization of low modulus and strong adhesion. At a fundamental level, silicone systems rely on covalent crosslinking or physical chain entanglement to balance mechanical compliance and interfacial conformity, resulting in an inherent trade-off between softness and adhesion. Moreover, low-molecular-weight siloxanes (D₃-D₁₀) must be reduced through repeated distillation, which increases manufacturing cost while failing to fully eliminate these species. Consequently, volatile siloxane release and oil bleeding remain persistent concerns. In previous work^[38-43], poly(ionic liquid) was shown to combine strong interfacial wetting and adhesion with highly designable ionic architectures, indicating their potential as replacements for silicone-based materials. In poly(ionic liquid) TIMs, ionic interactions can dynamically rupture and reform under strain, while localized stresses are effectively attenuated through ion migration. Consequently, the TIMs function as intrinsically stress-adaptive layers rather than purely passive load-bearing medium, showing a mechanical response challenging to achieve with conventional silicone-based TIMs.

Contact thermal resistance (R_c) represents the contact thermal resistance of the interface. The contact thermal resistances on both sides of the thermal interface material are R_c1 and R_c2 respectively, while thermal interface material resistance (R_TIM) represents the total thermal resistance of the thermal interface material. R_TIM = R_bulk + R_c1 + R_c2, R_bulk represents the intrinsic thermal resistance of the material. Currently, most research focus on the improvement of thermal conductivity of TIM, namely κ_TIM, thus lead to a reduced R_TIM by overlooking R_c1 and R_c2. The contact thermal resistances R_c1 and R_c2 are related to the surface roughness of the interface, the material thermal conductivity and hardness, the modulus of Young's, wettability and other properties^[44].

In this work, a solvent-free poly(ionic liquid) (PIL)-based thermal gel was presented. A dual-confinement network comprising ionic and covalent bonds, together with a dynamically reconfigurable ionic network, introduced an additional design parameter for tuning TIM properties, achieving the coexistence of low modulus and high adhesion performance. An ionic liquid monomer bearing vinyl groups was used as the polymerizable matrix, with Al₂O₃ particles as thermally conductive fillers. Upon mixing with an initiator, a dispensable formulation was obtained and then cured into a TIM at 85 °C for 3 h after lid attachment. To strengthen filler–matrix interactions, Al₂O₃ surfaces were prefunctionalized with vinyl groups, enabling covalent bonding with the PIL matrix. Thermally initiated polymerization of the vinyl-functional monomers allowed a one-step curing process, yielding a highly processable composite paste suitable for precise dispensing and controlled deposition (Scheme 1). At the Al₂O₃ loading of 70 vol%, the formulation exhibited a viscosity of 225 Pa·s at a shear rate of 5 s^-1, compatible with cartridge filling and standard dispensing processes. Compared with conventional silicone-based systems, the PIL gel cured at a lower temperature and was storable at room temperature without refrigeration. The intrinsic thermal resistance reached 2.4 × 10^-5 m²·K/W, comparable to that of silicone TIMs. Owing to synergistic electrostatic, hydrogen-bonding, and ion-dipole interactions at the interface, the cured TIM containing 70 vol% Al₂O₃ exhibited an interfacial adhesion strength of 0.91 MPa and an interfacial contact thermal resistance of R_c = 1.95 ± 0.87 × 10^-7 m²·K/W on Si substrates, substantially outperforming silicone counterparts. By comparison, a commercial silicone TIM (Shin-Etsu X-23-7772-4) showed an adhesion strength below 0.08 MPa and R_c = 2.61 ± 0.50 × 10^-6 m²·K/W on Si. The cured PIL gel exhibited a storage modulus of 255 kPa, comparable to typical silicone-based TIMs (~200 kPa). Reliability was evaluated using Si–TIM–Si dummy structures subjected to high-temperature storage (150 °C, 300 h), damp heat (85 °C/85% RH, 150 h), and thermal shock cycling (-50 to 150 °C, 30 min per cycle, 150 h). After testing, TIM coverage remained above 98%, indicating that the combination of low modulus and strong interfacial adhesion preserves interfacial conformity under service conditions. Departing from the conventional reliance on silicone or thermosetting matrices, this PIL-based thermal gel establishes a TIM platform that combines low modulus, strong interfacial adaptability, and high reliability, offering strong potential for high-heat-flux, large-scale advanced packaging in AI chips and post-Moore electronics.

Display Full Size

Scheme 1. Development and preparation of thermal gel for poly(ionic liquid) dispensing systems.

2. Methods

2.1 Materials

Sulfuryl chloride, triethylene, chloroform, trichloromethane, ethylene glycol monomethyl ether, pyridine, anhydrous magnesium sulfate, N-Vinylimidazole, ethyl ether, lithium bis(trifluoromethanesulfonyl)imide, ethanol, poly(ethylene glycol) diacrylate, and vinyl triethoxy silane were purchased from Millipore Chemicals. Spherical alumina was obtained from Jiangsu Lianrui New Materials Company.

2.2 Experimental details

2.2.1 Synthesis of VIm-3EO NTf₂

Mix 0.45 mol of sulfuryl chloride solvent with 90 mL of chloroform. Slowly add this mixture to the solution of triethylene glycol monomethyl ether (0.45 mol), pyridine (0.3 mol), and chloroform (200 mL) over a period of 60 minutes. Stir the resulting mixture in an oil bath at 100 °C for 4 hours, yielding a yellow turbid solution. Wash it four times with pure water, and dry over anhydrous MgSO₄. Perform vacuum distillation at 60 °C to remove chloroform. Purify the crude product under vacuum to obtain 2-[2-(2-methoxyethoxy)ethoxy]ethyl chloride, an orange transparent liquid.

2-[2-(2-methoxyethoxy)ethoxy]ethyl chloride and N-ethyleneimine were reacted at a molar ratio of 1:1 at 100 °C for 72 hours. After the reaction, the mixture was washed with ether to remove unreacted impurities and dried in an oven. Subsequently, it was subjected to ion exchange with lithium bis(trifluoromethanesulfonyl)imide in a water solution at a molar ratio of 1:1. Following the exchange, the mixture underwent liquid-liquid separation and drying to obtain the 1-ethylene-3-1-(2-(2-(2-ethoxyethoxy)ethoxy)ethyl)-3-ethylene-imidazole bis(trifluoromethanesulfonyl)imide salt product.

2.2.2 Preparation of modified Al₂O₃

The purchased Al₂O₃ powder was ultrasonically cleaned with anhydrous ethanol for 30 minutes to remove the adsorbed organic substances, and then dried in a vacuum at 150 °C for 12 hours to eliminate residual moisture. The vinyl triethoxy silane: water: ethanol volume ratio was set at 1:5:30 for mixing and the mixture was hydrolyzed at 60 °C for one hour. Then, the pre-treated activated Al₂O₃ powder was dispersed in the hydrolyzed mixed solvent, with a small amount of acetic acid added to maintain the weak acidity of the solvent. The reaction temperature was adjusted to 80 °C for 6 hours of mechanical stirring. Subsequently, the modified mixture was centrifuged to separate the activated Al₂O₃, washed with ethanol until neutral, and dried in a vacuum oven at 80 °C for 24 hours to yield the modified Al₂O₃ powder raw material.

2.2.3 Preparation of thermal gel

First, mix the crosslinking agent, initiator and ionic liquid monomer in a ratio of 1:1:200 to obtain the initial mixture. Then, respectively mix the modified Al₂O₃ at volume ratios of 50%, 60% and 70% with the mixture, and mix them in a vacuum mixer at a speed of 1,500 ppm for 2 minutes. Inject the uniformly mixed thermal gel into a syringe for further vacuum defoaming treatment. After that, inject it into the mold and place it in an oven for curing at 85 °C for 3 hours to obtain TIM.

2.3 Characterization

The ¹H NMR spectra were recorded on a Bruker 400 MHz spectrometer, using TMS as the internal reference. Attenuated total reflectance infrared (ATR-IR) spectra were recorded using a Fourier transform infrared (FTIR) spectrometer equipped with a diamond single-reflection ATR accessory. Thermogravimetric analysis (TGA) was conducted based on an SDT650 thermal analysis system (TA Instruments, USA) at a heating rate of 5 °C min^-1 under both air and N₂ atmospheres. Mechanical testing was carried out at room temperature using a universal tensile testing machine (AGX-10kNVD, Shimadzu) in accordance with ASTM D412. Dynamic viscoelastic properties of the composites were characterized using an Anton Paar MCR-302 rheometer. Frequency sweep tests were performed over an angular frequency (ω) range of 0.1-100 rad s^-1 at specified temperatures. Temperature sweep measurements were conducted from 25 to 180 °C at a heating rate of 5 °C/min, with a fixed angular frequency of 5 Hz and a shear strain (γ) of 1%. Adhesion strength was evaluated using a DAGE 4000HS high-speed multi-function weld strength tester. Dynamic mechanical analysis (DMA) and compressive stress–strain tests were performed using a TA Instruments RSA-G2 solids analyzer on square-shaped samples. In the DMA test, the composites were first compressed to 90% of their original thickness, and the compressive modulus was subsequently measured. They were then held for 15 min, followed by 15 min of stress relaxation, after which the rebound degree was measured. Sample morphologies and elemental distributions were analysed using scanning electron microscopy (SEM, Apreo 2S, Thermo Fisher Scientific) and energy-dispersive spectroscopy (EDS, Oxford Unity BEX). The samples after curing were examined by Micro CT using the Zeiss Xradia 620 Versa three-dimensional X-ray microscope. X-ray photoelectron spectroscopy (XPS) was performed using an ESCALA instrument (Thermo Fisher Scientific). The X-ray Cheetah-EVO equipment produced by German company YXLON was used to conduct flaw detection on sandwich structure samples. The structure of the sample was scanned with the American Sonoscan ultrasonic scanning microscope D9600 C-SAM. Through-plane thermal diffusivities of the PIL/Al₂O₃ composites were measured at 25 °C using the laser flash method with an LFA 467 instrument (NETZSCH, Germany). Thermal conductivity (k) was subsequently calculated using the following equation:

(1)$k=\alpha \cdot \rho \cdot C_p$

where α is the thermal diffusivity, ρ is the density, and C_p is the specific heat capacity of composites. Density was measured using the buoyancy method (SUNNY HENGPING, FA2104J, China), while C_p was determined by the sapphire method and calculated from differential scanning calorimetry (DSC) data. Thermal resistance and heat dissipation of the chip under practical conditions were simulated using a T3ster thermal resistance tester (Mentor Graphics). The total thermal resistance of the system is expressed by the formula: $R=\frac{T_2-T_1}{p}$, where R represents the thermal resistance, T₂ and T₁ are the chip’s temperature and the heat sink, respectively, and p is the output power of the chip. During the T3ster thermal resistance test, samples measuring 1 × 1 cm with a thickness of 1.5 mm were placed between the chip and heat sink. The chip temperature was recorded at power levels of 2, 4, 5, 6, 8, and 10 W. Thermal shock stability and durability were evaluated via cyclic heating/cooling tests at an applied power of approximately 5.0 W. Heat dissipation of the PIL/Al₂O₃ composite TIM was monitored using an infrared (IR) camera (FLUKE Ti480, USA).

2.4 Frequency Domain Thermoreflectance (FDTR) measurement

The configuration of the low-frequency FDTR system employed in this study is described as follows. A 532 nm green laser with a power of about 110 mW served as the probe beam, and a gold (Au) coating as the transducer. The pump source was a 457 nm blue laser delivering up to 1.5 W of energy, with roughly 60% of the incident power absorbed by the sample. Its modulation was controlled by a lock-in amplifier through an acousto-optic modulator, with a frequency ranging from 10 Hz to 100 kHz. Continuously adjustable neutral density (ND) filters were placed in both the pump and probe paths to precisely regulate laser power and maintain optimal measurement conditions. Prior to measurements, a sensitivity analysis was conducted to identify parameters measurable with high confidence and to select the appropriate modulation frequencies.

In FDTR, the periodically modulated pump beam induces a frequency-dependent surface temperature oscillation detected via probe-beam thermoreflectance. A lock-in amplifier extracts the phase lag ϕ, fitted with a multilayer heat-diffusion model over the frequency sweep to obtain thermophysical parameters. The thermal penetration depth (TPD) decreases with increasing modulation frequency and is commonly expressed as:

(2)$TPD=\sqrt{\frac{\alpha }{\pi f}}=\sqrt{\frac{k}{C\pi f}}$

where f is the modulation frequency, α is the thermal diffusivity, k is the thermal conductivity, and C is the volumetric heat capacity. The TPD is defined as the characteristic depth at which the amplitude of the temperature oscillation decays to 1/e of its surface value under periodic heating.

As the thermal penetration depth depends on modulation frequency, the phase sensitivity to each parameter also varies with frequency. In this study, the sensitivity of phase to a parameter x is defined as follows.

(3)$S_x=\frac{d\Phi }{d\ln x}$

The sensitivity curves are determined by the thermal model and are influenced by other inputs, such as laser spot sizes, sample structure, volumetric heat capacity, thermal conductivity, and contact resistance. When the sensitivity peaks of k and Rc do not overlap, multi-frequency fitting enables their effective decoupling. In our low-frequency FDTR results, the sensitivity peak of k occurs at lower frequencies than that of Rc.

3. Results and Discussion

As shown in Figure 1, the adhesive materials for the composites mainly consist of surface-modified alumina (Al₂O₃), VIm-3EO NTf₂, and poly(ethylene glycol) diacrylate (PEGDA) as the crosslinking agent. The vinyl-functionalized imidazolium ionic liquid monomer (VIm-3EO NTf₂) undergoes free-radical polymerization to form a PIL. The vinyl groups (-CH=CH₂) of the monomer participate in addition reactions to generate a polymer chain with repeating imidazolium NTf₂^- units. This PIL retains the chemical stability intrinsic to ionic liquid, imparting functional attributes to the composites. PEGDA, as a crosslinker, contains dual acrylate (-COOCH=CH₂) moieties at its chain termini. These acrylate groups copolymerize with the vinyl functionalities on both the surface-modified Al₂O₃ and the PIL chains where vinyl sites persist. This reaction constructs a three-dimensional crosslinked network where Al₂O₃ are covalently anchored within the polymer matrix. The PEGDA-derived network integrates the inorganic (Al₂O₃) and organic (PIL) phases, synergistically enhancing properties such as mechanical strength and thermal stability.

Display Full Size

Figure 1. Schematic diagram of the composite preparation process. Vinyl ionic liquid is blended with modified Al₂O₃, an appropriate crosslinking agent is introduced, and the composite is obtained after polymerization.

Figure 2a presents the ¹H nuclear magnetic resonance (¹H NMR) spectrum of the VIm-3EO NTf₂, where characteristic peaks (labeled a–g) correspond to distinct proton environments within the molecular framework. For instance: signals at ~5.0-6.0 ppm (Figure 2d,e) are attributed to vinyl (-CH=CH₂) protons, confirming the presence of polymerizable double bonds. Multiplets at ~7.0–10.0 ppm (Figure 2a,b,c) correlate with protons on the imidazolium ring, consistent with the designed molecular architecture. Upfield resonances (Figure 2f) correspond to alkyl chain protons, collectively validating the VIm-3EO NTf₂’s structural fidelity. Figure S1 shows the infrared spectrum of VIm-3EO NTf₂. The C=C stretching vibration of the vinyl group corresponds to the peak near 1,657 cm^-1 in the Figure. The conjugative vibration of C=N and C=C in the imidazole ring corresponds to the peaks near 1,571 cm^-1 and 1,553 cm^-1 in the spectrum. The unsaturated C-H stretching vibration of the imidazole ring corresponds to the peak near 3,149 cm^-1 in the spectrum. The C-O-C stretching vibration corresponds to the peak near 1,052 cm^-1. The saturated carbon of the alkyl C-H side chain corresponds to the peak near 2,883 cm^-1. This further verifies the structure of the ionic liquid. The particle size distribution of pristine Al₂O₃ nanoparticles (Figure S2) exhibits a unimodal profile with a peak centered at ~6.1 μm. This narrow dispersion indicates uniform primary particle dimensions, which is critical for achieving consistent surface modification and excellent fluidity after material mixing during the dispensing process. Fourier transform infrared (FT-IR) spectroscopy (Figure 2b) captures the chemical transformation of Al₂O₃ surfaces. For pristine Al₂O₃, broad O–H stretching (~3,400 cm^-1) and Al–O–Al vibrational modes (~500-1,000 cm^-1) dominate, reflecting abundant surface hydroxyl groups. After modification, two new absorption bands appeared at ~2,870-2,970 cm^-1 and ~900-950 cm^-1, which were attributed to the stretching vibrations of the Si-O-Al bond and -CH, respectively, providing direct evidence of covalent grafting of the silane modifier onto the Al₂O₃ surface. The Figure presents the Al₂O₃ powders before and after modification. In the optical photos, the alumina powder after modification is more uniformly dispersed compared to the unmodified powder, with no apparent agglomeration. SEM micrographs reveal morphological evolution: Pristine Al₂O₃ (Figure S3a) consists of agglomerated spherical particles, likely driven by interparticle hydrogen bonding of surface hydroxyl groups. Modified Al₂O₃ (Figure S3b) displays reduced agglomeration, as steric repulsion from grafted organic chains disrupts intermolecular interactions. Energy-dispersive X-ray spectroscopy (EDS) mapping confirms that uniform distribution of Al and O in both samples, verifying Al₂O₃ phase purity. To further evaluate the effect of different aluminum oxide volume fractions on the thermal gel, the Al₂O₃ was combined with the PIL in proportions of 50 vol%, 60 vol%, and 70 vol% to prepare the thermal gel. Figure 2c shows the heat flow curve during the curing process of the 70 vol% thermal gel. The result indicates that the material begins to cure at 85 °C, consistent with the activation temperature of the selected initiator. Figure 2d schematically delineates the sequential fabrication protocol of the composite, which involves three key steps: dispensing thermal gel, placing heat sink and curing process. Figure 2e shows that the thermal gel containing alumina volume fractions of 50 vol%, 60 vol% and 70 vol% had good fluidity before curing. In contrast, the 75 vol% thermal gel demonstrated a relatively high viscosity and limited flowability, while the 80 vol% mixture, after being mixed in the mixer, became spherical rather than in a flowing state. Therefore, this work focuses on 50 vol%, 60 vol% and 70 vol% of the thermal gel. This systematic workflow and excellent mixing effect ensure reproducible composite patterning. As validated by the real-time dispensing photograph in Figure 2f, the material exhibits controlled flow during extrusion, confirming processability for precise microscale patterning. The viscosity profiles (Figures 2g, before curing) reveal a pronounced shear-thinning behavior across all volume fractions (50 vol%, 60 vol%, and 70 vol%). Viscosity decreases with increasing shear rate (γ̇), enabling smooth flow during dispensing while resisting slumping at rest, which is critical for maintaining printed feature fidelity. The viscosity of the 70 vol% gel before curing was 225 Pa·s (at 5 s^-1). By comparing the infrared curves before and after solidification (Figure 2h), the conversion rate of the ionic liquid can be calculated using the following formula:

Display Full Size

Figure 2. (a) ¹H NMR spectrum (CDCl₃) of VIm-3EO NTf₂; (b) Infrared spectra of Al₂O₃ before and after modification; (c) The DSC heat flow curve during the gel curing process of 70 vol% filler; (d) Dispensing process schematic diagram; (e) Optical images of mixtures with different volume fractions; (f) Optical image of the dispensing of 70 vol% mixture; (g) The viscosities of mixtures with different volume fractions before curing; (h) 70 vol% thermal gel before and after curing's infrared spectrum. NMR: nuclear magnetic resonance; DSC: differential scanning calorimetry.

(4)$X=1-\left(\frac{C=C_{\text{vinyl}}^P/C=C_{\text{imidazole}}^P}{C=C_{\text{vinyl}}^m/C=C_{\text{imidazole}}^m}\right)\ast 100\%$

X represents the conversion rate of the sample, while $C=C_{\text{vinyl}}^P$ and $C=C_{imidazole}^P$ denote the peak areas of the infrared spectra of the vinyl group and imidazole ring in the PIL, and $C=C_{vinyl}^m$ and $C=C_{imidazole}^m$ represent the peak areas of the infrared spectra of the vinyl group and imidazole ring in the ionic liquid monomer. Based on the calculation, result shows that the monomer conversion rate is 65.9%. These results indicate that the ionic liquid component in the system is not completely polymerized into poly(ionic liquid), and the cured matrix therefore exists as a PIL-IL gel system. The residual monomers in the system do not affect the thermal stability of the material or its volatilization during use. This conclusion originates from the fact that the thermal decomposition temperature of the ionic liquid monomers is greater than 320 °C in the thermogravimetric curves (Figure S4). Through the leakage test, it can be seen that no significant leakage phenomenon occurs in the material (Figure S5). At the same time, the ionic liquid monomer can function as an internal plasticizer for the poly(ionic liquid) matrix, embedding itself between the polymer chains. Free IL small molecules are distributed between the polymer chain segments of the PIL macromolecules, weakening the Coulomb force and van der Waals force between the chains and thereby reducing the entanglement of the segments. The presence of the plasticizer increases the free volume within the PIL macromolecules, facilitating segmental mobility and lowering the glass transition temperature of the material. As a result, the system still retains good fluidity and flexibility. The reduced chain entanglement and intermolecular interactions decrease the overall viscosity of the system, enhancing the material’s flowability, processability, and viscoelasticity. This modification helps mitigate the brittleness typically associated with excessively high crosslinking density in poly(ionic liquid), making the material more compatible with common thermal management processing techniques such as dispensing and coating. Furthermore, it enables better thermal expansion matching with chip packaging, meeting the requirements for reliable heat dissipation.

The optical images in Figure 3a reveal the composite’s excellent mechanical flexibility under twisting and bending, even at high volume fractions. The composites maintain structural integrity under large deformations, indicating substantial ductility and strain tolerance, critical for applications in flexible electronics, wearable devices, and adaptive bonding systems. This flexibility originates from the well-integrated matrix-filler interface, where interfacial interactions effectively dissipate stress during deformation. Composites with volume fractions of 50% and 70% both exhibit stable deformability. This further indicates that they have good phase compatibility, enabling them to deform without fracturing under mechanical stress and to resist potential material failure caused by such stress. A Micro CT examination was conducted on the 70 vol% composite (Figure 3b), and the three-dimensional views of the composite were presented. The right figure shows the cross-sectional slice of the composite. It was found that the distribution of Al₂O₃ in the composite was uniform and relatively dense, further demonstrating the successful preparation of the compound. Figure 3c shows the cross-sectional morphology of the composites with different volume fractions. The density of the Al₂O₃ in the PIL matrix increases with its higher volume fractions, and no obvious holes or cracks are observed in the materials, indicating the completeness of the material preparation process. The EDS results (Figures S6,S7,S8) show that the spherical Al₂O₃ particles in the three volume fractions of the composites are uniformly distributed in the PIL matrix, indicating the uniformity of the material preparation. In order to analyze the XRD results, the crystal phase of Al₂O₃ powder was further determined and the phase stability of Al₂O₃ in different volume fractions (50%, 60%, and 70%) of the composites was verified (Figure S9).The XRD pattern of the pristine Powder Al₂O₃ exhibits sharp and intense diffraction peaks, which match the characteristic peaks of the α-Al₂O₃ phase. For the composites, their XRD patterns retain the characteristic diffraction peaks of α-Al₂O₃, this suggests that the crystal phase of Al₂O₃ remains unchanged during composite fabrication. Figure 3d schematically deciphers the tensile deformation pathway of the composite. Initially, Al₂O₃ particles are homogeneously dispersed within the PIL matrix, forming a dual-phase network where flexible PIL chains entangle around rigid inorganic fillers. Under tensile loading, PIL chains undergo elastic stretching and localized plastic slip, while interfacial interactions mediate load transfer between fillers and the matrix. As strain intensifies, two failure modes emerge: (i) interfacial debonding (inset, left), where filler–matrix contacts rupture due to shear stress; (ii) matrix fracture (inset, right), where overstretched polymer chains break. This microstructure-driven deformation logic directly dictates the composite’s strength and ductility-dependent mechanical behavior. The stress–strain profiles (Figure 4e) reveal a stark filler content-dependent mechanical dichotomy: 70 vol% Al₂O₃ exhibits the highest peak stress but the lowest fracture strain (43%), as dense filler packing constrains polymer chain mobility, amplifying load-bearing capacity while suppressing plastic deformation. 50 vol% Al₂O₃ displays maximal ductility (192%), with matrix-dominated deformation enabling extensive chain slippage and strain hardening. Figure 4f further quantifies modulus (black curve) and maximum tensile stress (σ_max, red curve) evolution. The maximum stresses at 50 vol%, 60 vol% and 70 vol% were 0.175 MPa, 0.327 MPa and 0.655 MPa respectively, and the modulus were 15.3 kPa, 183 kPa and 775 kPa respectively. Stress relaxation tests were conducted on three different volume fractions of the composites (constant 10% strain, DMA compression 15 minutes relaxation). During the relaxation process, the stress decreased gradually leveled off. The modulus of the 70 vol% composites was approximately ten times that of the 50 vol% composites (Figure S10). After 15 minutes of stress relaxation, the composites with different filler contents showed varying degrees of rebound. The composite materials with 50%, 60%, and 70% filler content rebounded to 98.2%, 97.3%, and 96.6% of the initial thickness, respectively, which confirmed that the Al₂O₃ played a mechanical anchoring role, delaying the viscoelastic decay and enhancing the temporal stability of the structure. The variation of material thickness with temperature is shown in Figure S11. The thermal expansion coefficient of the material decreases as the volume fraction of the filler increases. The coefficients of the composites with 50%, 60% and 70% filler content at 80 °C are 239 ppm/°C, 127 ppm/°C and 88 ppm/°C respectively. Thermogravimetric analysis (TGA) can clarify the thermal degradation behavior under inert atmosphere (Figure 4g) and in air environment (Figure S12). Due to the excellent thermal stability of Al₂O₃, there is less mass loss as the content of Al₂O₃ increases. Due to the good thermal stability of PIL, samples with different contents of Al₂O₃ all have higher thermal decomposition temperatures. The temperature at which 5% of the mass is lost is defined as the thermal degradation temperature. The thermal degradation temperatures at 50, 60, and 70 vol% are 402, 411, and 426 °C in nitrogen, respectively, and 398, 407, and 414 °C in air. It is worth noting that the high thermal conductivity of Al₂O₃ can dissipate heat, while its chemical inertness inhibits the reaction between the matrix and oxygen, thereby delaying decomposition. In addition, the TGA curves of composites with different volume fractions were tested under both air and N₂ atmospheres at a temperature of 150 °C for 300 minutes (Figure S13). The results showed that the mass of the composite materials did not significantly decrease under long-term high-temperature conditions, indicating the good thermal stability of the materials. Meanwhile, due to the self-healing property of PIL, the PIL/Al₂O₃ composite material also exhibits significant self-repairing characteristics after being heated to 80 ℃ for 5 hours, as shown in Figure 3h.

Display Full Size

Figure 3. (a) Optical images of the composites with volume fractions of 50% and 70% under twisting and bending deformations; (b) Micro CT image of 70 vol% composite; (c) SEM images of composites with different Al₂O₃ contents and EDS of composite; (d) Schematic diagram of tensile of composites; (e) Tensile curves of composites with different volume fractions and (f) the corresponding modulus and maximum stresses; (g)Thermogravimetric curves of composites under N₂ atmosphere; (h) Composites after being cut and then they undergoing self-healing optical imaging. SEM: scanning electron microscopy; CT: computed tomography; EDS: energy-dispersive spectroscopy.

Display Full Size

Figure 4. (a) The viscosities of mixtures with different volume fractions after curing; (b) the frequency dependence and (c) temperature dependence of the storage modulus and loss modulus; (d) Schematic of adhesion; (e) Optical demonstrations of the composite’s adhesion to diverse substrates and the load-bearing capacity of Glass-Steel and Glass-Cu bonded joints; (f) Schematic illustration of the multi-mechanism adhesion between the composite and substrates; (g) Adhesion strength of the composite with different volume fractions at room temperature on various substrates.

Notably, cured composites exhibit orders-of-magnitude higher viscosity than their uncured counterparts (Figure 4a), attributed to the formation of a crosslinking-induced network that restricts molecular mobility. This effect becomes more pronounced with higher volume fractions due to enhanced particle–matrix interactions and physical entanglement. The complex viscosity (η*) (Figure S14) mirrors the shear viscosity trends: η* increases with volume fraction and decreases with angular frequency (ω), consistent with shear-thinning behavior. Frequency sweeps (Figure 3b) reveal that the storage modulus (G’), representing elastic energy storage, consistently exceeds the loss modulus (G’’) (viscous energy dissipation) across the tested range. This G’ > G’’ relationship confirms a predominantly elastic, gel-like network, which is essential for structural integrity post-dispensing. At higher volume fractions (e.g., 70 vol%), G’ rises more steeply with ω, indicative of enhanced network rigidity from denser particle packing and crosslinking. Temperature-dependent modulus profiles (Figure 3c) further demonstrate the composite’s thermal robustness, as G’ remains superior to G’’ even at elevated temperatures (up to 180 °C).

The thermal gel used in this work has excellent adhesion properties (Figure 4d). These properties enhance the ability of the TIM to maintain good interfacial contact under complex thermal and mechanical stresses in actual working conditions. Figure 4e demonstrates the composite’s broad substrate compatibility, displaying robust bonding to inorganic (Si, Glass, Metal) and organic (PP) surfaces. The successful adhesion to chemically distinct substrates highlights the composite’s adaptability to diverse surface chemistries. Notably, the Glass-Steel and Glass-Cu joints support a 500 g load, experimentally validating the composite’s practical adhesive capacity. The multi-mechanism adhesion schematic (Figure 4f) explains the molecular origins of strong adhesion. Metal complexation occurs through the formation of coordinate bonds between metal cations (e.g., Fe²⁺, Al³⁺, Cu²⁺) on metallic substrates and anionic groups (e.g., O^2-, Si-O^-) in the composite. Electrostatic interactions arise from attractive forces between charged species on the composite and substrate surfaces. Hydrogen bonding further contributes to adhesion via strong interactions between hydroxyl (-OH) groups (abundant on Glass/Si) and polar functional groups in the composite. Ion-dipole/dipole-dipole interactions provide long-range forces between ionic species/dipoles in the composite and substrate. Van der Waals interactions, as universal intermolecular forces, enhance interfacial adhesion. This synergistic combination of specific (e.g., hydrogen bonding, metal complexation) and nonspecific (van der Waals) interactions forms a robust bonding network, accounting for the composite’s strong and durable adhesion. Figure 4e quantifies adhesion strength across substrates and volume fractions. Glass exhibits the highest adhesion strength, with values of 1.85 ± 0.21 MPa, 1.58 ± 0.15 MPa and 1.01 ± 0.11 MPa for 50, 60, and 70 vol%, respectively. This is attributed to its dense -OH groups that form extensive hydrogen bonds with the composites. Al, Si, and Cu show moderate yet still substantial strengths. The 50 vol% composite generally outperforms 60 vol% and 70 vol% counterparts. At 70 vol%, excessive filler may induce agglomeration, reducing effective interfacial contact and weakening matrix connectivity before lowering adhesion. Figure S15 presents the interfacial adhesion strength of the 70 vol% thermal gel during the test at different temperatures. The results indicate that as the temperature increases, the interfacial adhesion strength decreases significantly at 50 °C, but remains above 0.1 MPa at temperatures below 175 °C, maintaining a weak interfacial adhesion strength. The main reasons for the attenuation of the interface adhesion strength of composites under high temperature may be as follows: (1) The T_g of PILs decreases under high temperature, and the chain segments undergo viscous flow. Under thermal stress, relative slippage at the interface may occur, and macroscopically, the interface shear strength fails. (2) High temperature intensifies the mobility of PIL chain segments, disrupting the polar matching between the ionic groups and the polar surfaces of inorganic/organic fillers, thus resulting in a sharp decline in interface compatibility.

Thermal management performance under practical operating conditions was evaluated using a T3ster system (Figure 5a). A constant power was applied to heat the ceramic sheet, and the prepared samples were used to fill in the space between the heat sink and the heated ceramic sheet to promote heat transfer. Figure 5b depicts the temperature evolution for samples without TIM and with TIMs at different volume fractions: the sample without TIM exhibits a rapid temperature rise to over 120 °C, whereas TIM-integrated samples show markedly lower temperatures, with the 70 vol% TIM sample maintaining the lowest temperature. This demonstrates that the composite TIM effectively accelerates heat dissipation from the heat source. The steady-state temperature as a function of power density (Figure 5c) further quantifies this performance: as power density increases, samples with higher filler contents exhibit a significantly slower temperature rise, confirming their superior capability to manage high heat fluxes.

Display Full Size

Figure 5. Thermal management performance of composites with varying filler volume fractions. (a) The schematic diagram of the T3ster system and the heat transfer diagram of the thermal gel. The temperature of the chip varies with the (b)working time and (c) different input powers; (d)The sensitivity analysis of the FDTR test and the (e) test results of 70 vol% TIM; (f) The results of 1,000 heating-cooling temperature cycles of 70% TIM solution; (g) The thermal infrared images of different filler contents TIM on the heating device and (h) changes in surface temperature over time. FDTR: Frequency Domain Thermoreflectance; TIM: thermal interface material.

Thermal conductivity and thermal diffusivity are two fundamental thermal properties that control the steady-state and transient response of materials under thermal load. Accurate measurement of these parameters is critically important, and is affected by many factors, such as the physical characteristics of the sample, ambient temperature, and its intrinsic thermal conductivity and diffusivity. Although steady-state methods can measure thermal conductivity by applying a constant heat flux and measuring the corresponding temperature gradient, these approaches are usually laborious and require precise control of heating and boundary conditions. In contrast, transient techniques, which record the time-domain or frequency-domain response of materials under pulsed or periodic thermal excitation, provide a simpler and more efficient way to determine thermal conductivity and thermal diffusivity rapidly and reliably^[45-47]. To further elucidate the thermal transport mechanisms, FDTR measurements were conducted. The sample structure for the FDTR test and the corresponding schematic diagram are shown in Figures S16,S17. The sensitivity analysis (Figure 5d) illustrates the frequency dependence of the sensitivities to interface thermal resistance (R_c) and thermal conductivity (κ_TIM), revealing that FDTR can resolve their respective contributions within the tested frequency range (10¹-10⁶ Hz). The experimental FDTR results (Figure 5e) yield = 1.65 ± 0.16 W/m·K and R_c = 1.95 ± 0.87 × 10^-7 m²·K/W, which are consistent with the bulk thermal conductivity in Figure S18. The BLT in this work is approximately 40 μm, and according to the formula $R=\frac{BLT}{k}$, the bulk thermal resistance is calculated to be roughly 2.4 × 10^-5 m²·K/W. The reason why the thermal resistance at the material interface is much lower than that of the material body is the PIL-based TIM achieved a lower interface thermal resistance due to the unique interface bonding mechanism and good mechanical flexibility of PIL. However, in this study, since Al₂O₃ was used as the thermal conductive filler and the filler content was not high enough, the intrinsic thermal resistance of the material was relatively large. This results in the interface thermal resistance of the material being one order of magnitude lower than the bulk thermal resistance. Therefore, the future development direction of PIL-based TIM is to reduce the bulk thermal resistance.

A self-consistent test by using the FDTR method used in this paper to test both PIL/Al₂O₃ samples and the commercial TIM. For comparison, the interface thermal resistance of the commercial thermal gel was tested using the same standard, with the results shown in Figures S19,S20. The interface thermal resistance of the commercial gel is 2.61 ± 0.50 × 10^-6 m²·K/W, which means that the interface thermal resistance of the thermal gel in this study is approximately one order of magnitude lower than that of the commercial gel. Table 1 compares the BLT and R_c values of typical thermal interface materials, demonstrating the significant advantages of PIL/Al₂O₃.

This consistency validates both the reliability of the FDTR technique and the composite’s balanced intrinsic thermal conductivity and interface thermal resistance. Thermal cycling stability is critical for long-term reliability. As shown in Figure 5f, the 70 vol% TIM sample undergoes 1,000 heating–cooling cycles without noticeable degradation in its temperature profile, indicating excellent thermal cycle stability. This robustness ensures the composite’s reliability during prolonged operation under periodic heat loads. Infrared (IR) thermal imaging (Figure 5g,h) provides direct visualization of temperature distribution and its temporal evolution. Figure 5g captures the surface temperature profiles at different times for composites with 50 vol%, 60 vol%, and 70 vol% fillers. As the filler content increases, the temperature rises more rapidly and reaches a slightly higher steady-state value. The final steady-state temperature is 93.1 °C for 70 vol% and 91.9 °C for 50%, with more uniform heat distribution. The temperature–time curves derived from IR imaging (Figure 5h) quantitatively corroborate these observations, with the 70 vol% TIM sample exhibiting the fastest heating rate and highest equilibrium temperature. These results align well with T3ster measurements, collectively verifying the composite’s enhanced thermal management performance from both macroscopic temperature monitoring and spatial thermal distribution perspectives.

To further verify the reliability and industrial applicability of the samples, 70 vol% TIM was used to fabricate sandwich-structured samples. The uncured sample was drop-cast onto a silicon wafer, subsequently covered with a 1 × 1 cm silicon wafer, and then cured (Figure 6a). To comprehensively validate the material reliability, three aging tests were designed: high-temperature storage test (the cured samples were stored in an oven at 150 °C for 300 hours); high-temperature and high-humidity test (the samples were exposed to an environment of 85 °C and 85% relative humidity for 150 hours); thermal shock test (the samples were alternately incubated at -50 °C and 150 °C for 30 min per cycle and a total duration of 150 hours). After accelerated aging, the samples were subjected to X-ray scan, interfacial adhesion strength, and ultrasonic scanning tests, respectively. The test results are presented in Figure 6b,c. Interfacial adhesion strength tests on the aged samples revealed that, even after 300 hours or 150 hours of accelerated aging (Figure 6b), the interfacial adhesion strength decreased by less than 0.1 MPa, further confirming the excellent aging resistance of the samples. X-ray scan results (Figure 6c) indicated that no obvious internal voids or delamination were observed in the samples after the three aging tests, demonstrating that the samples maintained good stability under all accelerated aging conditions. In addition, an ultrasonic scanning microscope was employed to supplement the ultrasonic scanning results obtained after the aging experiment. The results showed that there was almost no delamination at the interface layer of the samples. The average coverage rate after the three aging experiments, quantified using the Image J calculation software, exceeded 98%. This confirmed that these samples maintained a high coverage rate even after accelerated aging. These results collectively indicated that the PIL-based thermal gel boasts good reliability and can meet the basic requirements for practical applications.

Display Full Size

Figure 6. (a) Schematic diagram of the sandwich sample structure preparation; (b)The interfacial adhesion strength obtained from the three accelerated aging experiments; (c) Images of the sample after undergoing three accelerated aging experiments taken by X-ray and the ultrasonic scanning images of the three aging samples after the longest aging period.

4. Conclusions

In conclusion, this study presents a novel thermal gel based on a PIL matrix and filled with Al₂O₃ particles, offering a new material strategy for high-performance thermal interface materials. The PIL matrix possesses inherent ionic cross-linking and high flexibility. The addition of the Al₂O₃ filler effectively constructs a continuous thermal conduction path, significantly improving the thermal conductivity while maintaining mechanical flexibility and adhesion strength (0.95 MPa with Cu and 0.91 MPa with Si). Comprehensive characterization has demonstrated the excellent performance of the PIL/Al₂O₃ thermal gel in rheological properties suitable for dispensing applications (225 Pa·s), thermal stability, low interface thermal resistance (R_c = 1.95 ± 0.87 × 10^-7 m²·K/W), mechanical properties, and anti-aging performance (the coverage rates after the three aging tests were all above 98%.). This work not only validates the feasibility of PIL as a new-generation matrix for TIM, but also establishes a molecular design framework that balances thermal and mechanical properties by integrating ionic interactions and filler engineering. These results provide valuable insights into the development of TIM1 such as thermal gel for next-generation electronic devices. Although the overall performance of the PIL-based thermally gel has not yet surpassed that of conventional silicone systems, this work validates the feasibility of this technological approach and highlights its potential for practical applications. Future research will focus on molecular structure optimization and the incorporation of functional additives to further enhance material properties and explore the performance limits of polymer-based TIMs.

Supplementary materials

The supplementary material for this article is available at: Supplementary materials.

Authors contribution

Zeng J: Data curation, formal analysis, investigation, writing-original draft.

Rao T: Formal analysis, visualization, methodology.

Shi H, Guo Y: Formal analysis, methodology.

Peng Z, Li L: Resources, supervision, validation.

Sun R: Supervision, funding acquisition, project administration, validation.

Yao Y: Conceptualization, funding acquisition, project administration, writing-original draft, writing-review & editing.

Conflicts of interest

The authors declare no potential conflict of interest exists.

Ethical approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Availability of data and materials

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Funding

This work was supported by National Natural Science Foundation of China (No. 62474116), CAS Proiect for Young Scientists in Basic Research (No. YSBR-105), National Key R & D Program of China (No. 2022YFA1203100), Autonomous deployment project of China National Key Laboratory of Materials for Integrated Circuits (No. SKLJC-Z2025-A08), GuangDong Basic and Applied Basic Research Foundation (No. 2025A1515012967), and Beijing Natural Science Foundation (No. L257012.

Copyright

© The Author(s) 2026.