High‐Yield‐Stress Particle‐Stabilized Emulsion for Form‐Factor‐Free Thermal Pastes with High Thermal Conductivity, Stability, and Recyclability

Thermal pastes, thermally conductive fillers dispersed in liquid matrices, are widely used as thermal interface materials (TIMs). TIMs transfer heat generated from electronics to the surroundings, ensuring optimal operating temperatures. Thus, it is crucial to obtain high thermal conductivity (TC) by forming a continuous heat‐conduction pathway through interconnected filler‐networks within the TIM. Therefore, for paste‐type TIMs with spherical fillers, high TC can only be realized at sufficiently high filler loadings (>60 vol%). However, the pastes bearing such high filler loadings are thick, stiff, and less applicable. To these ends, particle‐stabilized emulsions composed of immiscible liquids (silicone oil and glycerol) and spherical alumina are utilized as thermal pastes. Owing to this structure, the resulting form‐factor‐free thermal paste exhibits higher TC and stability than a simple mixture consisting of alumina and a single‐liquid‐matrix (either silicone oil or glycerol). Furthermore, the high applicability of the emulsion‐type pastes enables syringe extrusion, 3D printing, multiple cycles of reprocessing/molding, and eco‐friendly recycling.


Introduction
Thermal interface materials (or TIMs) are widely used to dissipate heat generated from an electronic component (or heat sources) to its heat sink. [1,2]The main advantage of TIMs is DOI: 10.1002/admi.202300860their ability to reduce the temperature of electronics during operation. [3]herefore, TIMs not only improve the performance and lifespan of electronics but also prevent devices from exploding and other possible risks caused by overheating. [4][7][8] Therefore, paste-type TIMs are widely utilized in electronic devices, light-emitting diodes, solar panels, and medical devices, particularly those with complex heat source structures.Solid and/or other types of TIMs with poor applicability, on the other hand, result in micro/macroscopic voids after application that can harbor trapped air. [9]This air acts as an insulator and significantly hinders heat transfer, causing the component to overheat and even heat-management failure. [10,11]lthough the exact composition of thermal paste can vary depending on the intended application, most thermal pastes for electronics consist of silicone oil and ceramic fillers. [12]Silicone oil is often used as a liquid matrix for the thermal paste owing to its good wettability, high thermal stability, and high chemical resistance. [12]Ceramics are ideal filler materials for TIMs because they exhibit high thermal conductivity (TC) while being electrically insulating. [13]Ensuring that TIMs are electrically insulating is crucial in electronics applications to reduce the risk of electrical shorts. [14]Furthermore, the excellent chemical stability and wear resistance of the ceramic fillers makes them suitable materials for protecting and cooling electronic components over consecutive heating/cooling cycles.In fact, spherical alumina (Al 2 O 3 ) microparticles are the most widely incorporated ceramic filler in industry, largely owing to their low cost coupled with the aforementioned advantages of ceramics. [13,15,16]owever, it is nearly impossible to maintain the long-term dispersion stability of spherical alumina microparticles in silicone oil because gravitational sedimentation dominates over the Brownian motion of the particles.Gravitational sedimentation leads to phase separation over time, resulting in a particle-enriched bottom layer and a particle-free top layer of silicon.[19] Several approaches, including particle surface modifications or the addition of stabilizing agents, have been developed to enhance particledispersion stability. [20,21]However, although these approaches improve the interfacial compatibility between the particles and liquid medium to a certain degree, they do not guarantee longterm sedimentation stability owing to the persistence of the gravitational body force.Moreover, the particle TC significantly deteriorated upon particle surface modification owing to the insulating layer formed between the conductive particles. [22]his instability issue may in part be mitigated by incorporating a high volume of filler into the liquid matrix, which dramatically increases the suspension viscosity and thereby imparts a strong resistance to the gravitational particle settling.The divergence of the suspension viscosity often accompanies the creation and/or increase in the "yield stress," attributed to the formation of the interconnected filler-network. [23,24]][27][28][29][30][31] Consequently, high TC and long-term dispersion stability can simultaneously be accomplished by increasing the filler loading to the maximum packing limit. [29]In reality, however, owing to the intrinsically low TC of silicone oil ≈0.18 W m −1 K) and the difficulty of achieving a seamless percolation network of the fillers, it is often challenging to achieve a TC above 2 W m −1 K for a thermal paste containing only alumina, even at very high filler loadings (>60 vol%). [9,16,27]Furthermore, at such high filler loadings, the thermal pastes become heavy and sometimes extremely vitrified, losing their applicability to devices of diverse geometry. [32]onstructing an effective filler-network at low filler loadings would thus be an ideal mean to obtain a thermal paste with high TC and long-term dispersion stability while maintaining its compliance with the desired device.[35] The dispersed droplets would basically occupy the given volume fraction in the continuous liquid medium, in-creasing the effective spatial packing density of the co-suspended fillers and restricting their translational motion.If fillers have a sufficiently small size, characterized by a small bond (Bo) number, and suitable wettability, they may even self-assemble at the interfaces between the dispersed droplets and the continuous liquid medium. [36][44] Using high-TC fillers, such properties may be exploited to enhance the TC of the resulting colloidal system because these interconnected fillers would form an effective thermal pathway in the insulating liquid matrix.
In this study, to these end, thermal pastes in the form of particle-stabilized emulsions were formulated via the simple combination of commonly used cost-effective ingredients, namely, silicone oil, glycerol, and spherical alumina.Using relatively small-sized (3 μm) and relatively large-sized (90 μm) fillers, Pickering-and oil-in-suspension-type emulsions were acquired, respectively.The thermal pastes of both emulsion-types were more thermally conductive and stable than simple suspensiontype thermal pastes consisting of a single-component liquid matrix and fillers in the corresponding volume fractions.These superior properties are attributed to the percolation facilitated by the formation of the filler-network, which will be discussed in terms of the emulsion yield stress measured using shear rheometry and the internal microstructure observed using confocal laser scanning microscopy.The high yield stresses of these particlestabilized emulsions enabled their versatile rheological processing, making them practically form-factor-free.Other advantages of these emulsions, such as heat and moisture stability along with recyclability, are also demonstrated, highlighting their versatility in various applications.

TC of the Thermal Pastes
Figure 1a illustrates the manufacturing processes of two types of thermal pastes: the single-liquid-matrix version (referred to as "suspension-type") and the two-liquid-matrix version (referred to as "emulsion-type").For the suspension-type thermal pastes, a blend of silicone oil or glycerol matrix and alumina was prepared by vigorously mixing followed by defoaming to ensure a uniformly dispersed state.In the case of emulsion-type thermal pastes, equal proportions of immiscible silicone oil and glycerol were vigorously mixed and defoamed.Then, alumina was introduced into the emulsion.Subsequently, thorough mixing and defoaming were carried out, resulting in stabilized emulsion-type thermal pastes.Alumina microparticles with average diameters of 3 and 90 μm were employed to assess the impact of filler size.All mixing and defoaming steps were conducted with the planetary centrifugal mixer.The square and triangle symbols represent thermal pastes containing alumina microparticles with average diameters of 3 and 90 μm, respectively.The theoretical TC as a function of alumina loading was predicted using the Bruggeman model (dashed line), as shown in Equation (1).In this prediction, the filler TC was held constant at 30 W m −1 K, the TC value of alumina reported in the literature. [13]e TC of thermal pastes as a function of filler loading is shown in Figure 1b,c, and the corresponding values are listed in Table S1 (Supporting Information).SA and GA denote the suspension-type thermal paste composed of silicone oil/alumina and glycerol/alumina, while SGA represents the emulsion-type thermal paste composed of a mixture of silicone oil/glycerol in a 1:1 volume ratio and alumina microparticles.The numerical suffixes appended to the sample labels indicate the average diameter of the alumina used.As shown in Figure 1b,c, a consistent upward trend in TC with increasing filler loading was observed, regardless of the matrix type or filler size.GA always exhibited a higher TC than SA.In SGA with an intermediate TC of the matrix, at relatively low filler loadings (<30 vol%), the TC values were intermediate to those of SA and GA.These findings confirm that a high-TC matrix induces a high TC of the composite. [45]Interestingly, at relatively high filler loadings (>30 vol%), the TC of SGA90 was comparable to that of GA90, while the TC of SGA03 surpassed that of GA03, which has the highest TC of the matrices studied here.
In order to investigate the mechanisms of the high-TC emulsion-type thermal pastes, the well-known Bruggeman model [46] was employed to fit the experimental TC of SA, GA, and SGA (Figure 1d-f).Among the many theoretical models for estimating the composite TC, the Bruggeman model, as expressed in Equation ( 1), provides accurate predictions, assuming that the composites have uniformly dispersed spherical fillers. [46,47] where  c ,  f , and  p are the TC of the composite, filler, and matrix, respectively; and V f is the filler volume fraction.To evaluate the effect of the matrix TC on the TC of the thermal paste, the filler TC was fixed to 30 W m −1 K, regardless of alumina size.This value corresponds to the reported TC of alumina in the literature. [13]As shown in Figure 1d,e, for suspension-type thermal pastes, the experimental TC values of SA03, GA03 (squares) and SA90, GA90 (triangles) were in relatively good agreement with the theoretically predicted values, indicating that the filler was uniformly dispersed in the matrix.On the contrary, as shown in Figure 1f, for emulsion-type thermal pastes, both SGA03 and SGA90 exhibited a higher experimental TC than the predicted values.This result implies that the stronger and/or more extensive the fillernetwork structural formation in the emulsion-type thermal paste, the more likely the model is to predict a composite with randomly and well-dispersed fillers.

Yield Stress of the Thermal Pastes
The yield stress values of the thermal pastes may reflect the relative significance of filler-network formation in different formulations.Yield stress, which likely originates from the formation of an internal structure within the fluid, is defined as the minimum stress required to plastically deform materials by breaking the internal microstructure.The yield stress of each thermal paste was estimated by fitting the complex viscosity as a function of angular frequency (Figure S1, Supporting Information) to the Herschel-Bulkley (H-B) model (a detailed description of this procedure can be found in the Experimental Section). [48]The measured yield stresses of the thermal pastes are illustrated in   S2, Supporting Information).Note that pure silicone oil and glycerol showed Newtonian behavior, where the viscosity is a function of angular frequency, resulting in virtually no yield stress.On the contrary, the mixture of silicone oil and glycerol (unstable emulsion form) exhibited a yielding behavior owing to the restricted translational and rotational motion of the dispersed droplets (Figure 2a,b; Figure S1, Supporting Information).As depicted in Figure 2a,b, the yield stress increased as the filler loading was increased for both the suspension-and emulsion-type thermal pastes, owing to the matrix-filler and/or filler-filler interactions.The emulsion-type thermal pastes, however, always exhibited higher yield stresses than the suspensiontype thermal pastes.For both SGA03 and SGA90, remarkable increments in yield stress were noted within the 0-10 vol% range of alumina loading.Such significant enhancements in yield stress were not detected in the suspension-type thermal pastes.These observations collectively suggest the establishment of a more robust filler-network structure within the SGA samples, even at alumina loadings below 10 vol%.The influence of greater yield stress in emulsion-type thermal pastes can be further substantiated through a comparison of particle stability against gravitational sedimentation.Figure 2c shows photographs of SA03 and SGA03 after 7 days of resting at room temperature.Figure S2 (Supporting Information) shows the other suspension-and emulsion-type thermal pastes with varying matrix and alumina sizes.As shown in Figure 2c and Figure S2 (Supporting Information), for all suspension-type thermal pastes, the alumina particles eventually settled to the bottom owing to gravity, even at room temperature.However, it is noteworthy that such gravitational sedimentation was not observed for the emulsion-type SGA thermal pastes, particularly those with alumina loadings greater than 30 vol%.Furthermore, as shown in Figure 2d, among the suspension-type thermal pastes, SA and GA containing 50 vol% of alumina (a relatively high loading) flowed immediately upon tilting.However, the emulsion-type SGA thermal paste containing 50 vol% alumina did not flow at all, retaining its shape for at least 7 days at room temperature.The remarkably higher sedimentation stability and shape retention of the emulsion-type thermal pastes are attributed to the stronger interfiller-network, as quantitatively demonstrated by their high yield stresses.

Correlation Between TC and Yield Stress of the Thermal Pastes
][51] This is not surprising because both parameters should be proportional to the degree of internal filler-network formation.For example, Prasher et al. suggested an empirical correlation between TC and yield stress of the TIM applied between the two solid surfaces, as expressed in Equation ( 2). [49] where k TIM , R bulk and  y are the TC, bulk thermal resistance and yield stress of the TIM, respectively; P is the pressure required to form a conformal coating of the thermal paste between two solid surfaces; and C and m are fitting constants.Under constant R bulk and P for a given paste formulation, the TC of the thermal paste should be directly proportional to its yield stress to the power of m.Here, a greater m value means that the relative enhancement in TC is greater than the relative enhancement in yield stress upon increasing the filler loading, and vice versa.This correlation has been developed for suspension-type TIMs with a single-component liquid matrix, but we adopted it in a heuristic fashion to investigate the properties of the newly developed emulsion-type thermal pastes.Figure 3a illustrates the correlations between the TC and yield stress of different paste formulations.It is noteworthy that an inflection point occurs at an alumina loading of ≈40 vol%, below and above which two distinct TC- y correlations were found.The extracted correlation factor m for different systems in the two regions is represented in Figure 3b (more detailed fitting results are shown in Table S3, Figure S3, Supporting Information).For alumina loadings below 40 vol%, the emulsion-type thermal pastes (SGA) exhibit higher m values than the suspension-type thermal pastes (SA and GA), suggesting that the TC increased more efficiently in the former case, attributed to the more effective fillernetwork formation.For alumina loadings above 40 vol%, the m values of all the thermal pastes are lower than those for loadings below 40 vol%, probably owing to the predominance of rheological percolation over thermal percolation for such high particle loadings.Note that this trend was particularly strong in the emulsion-type thermal pastes, supporting the more favored fillernetwork formation in those systems.
To corroborate the correlation between TC and yield stress, the internal morphology of both the suspension-and emulsion-type thermal pastes was directly observed using a confocal laser scan- Silicone oil (green) and glycerol (red) were stained with Nile red and Nile blue dyes, respectively, to clearly identify the microstructure of the thermal paste.As shown in Figure 4, the emulsiontype thermal pastes generally exhibited more ordered filler structures than the suspension-type thermal pastes such as GA03 and GA90, shown in Figures S4 and S5, (Supporting Information), respectively.As can be seen in the confocal micrograph of SGA03_10 (The leftmost image in Figure 4a), the black lines representing alumina particles are exclusively positioned at the interface between the uniformly sized silicone oil droplets and glycerol.This unequivocally signifies the formation of a Pickering emulsion. [37]With the incorporation of additional alumina (or a reduction in relative matrix volume), the size of the oil droplets gradually diminishes, and the fillers begin to interconnect with those adsorbed in neighboring silicone oil droplets, forming a progressively interconnected network structure (SGA03_20 in Figure 4a).At alumina loadings greater than 30 vol%, the size of the oil droplets in glycerol gradually decreased, and alumina began forming densely packed/bridged structures around silicone oil droplets, thereby forming the segregated filler structure.This formation of a segregated filler structure was responsible for the notable rise in both TC and yield stress observed in SGA03 in comparison to the corresponding suspension-type thermal pastes, as depicted in Figures 1b and 2a, respectively.
Owing to the comparatively larger size of the utilized alumina particles, however, achieving stability at the interface between silicone oil and glycerol could pose challenges, specifically for SGA90.Consequently, the dispersed oil droplets and alumina particles were randomly distributed within the continuous glycerol phase, unlike the distinct interfacial arrangement observed in SGA03 (see Figure 4a,b for SGA03 and Figure 4c,d for SGA90).Thus, SGA90 appears to be a so-called "oil-insuspension" emulsion. [38,39]As the filler loading increased up to 30 vol%, the alumina did not adsorb around the silicone oil droplets; instead, a thin layer of glycerol and microdroplets of glycerol/silicone oil enveloped the alumina.These insulating layers surrounding the alumina hindered the formation of the particle network, preventing efficient heat transfer.In fact, despite the lower interfacial thermal resistance owing to the larger size of alumina, [52] SGA90 exhibited a similar or slightly higher TC than SA90 at filler loadings below 40 vol%, even though the latter possesses the lowest matrix TC (Figure 1c).At filler loadings above 40 vol%, however, oil microdroplets trapped between the alumina particles contributed to improving the packing density of the alumina network.As a consequence, the TC of SGA90 surpassed that of GA90, despite the latter having the highest matrix TC (Figure 1c).Furthermore, irrespective of the filler loading, SGA90 consistently exhibited a higher yield stress than the suspensionlike thermal pastes.This phenomenon can be attributed to the random dispersion and entrapment of silicone oil microdroplets, which to some degree constrains the movement of the filler.However, it is important to note that the yield stress of SGA90 remains lower than that of SGA03, wherein a segregated filler structure is induced through Pickering emulsion (Figure 2a,b).
To summarize the aforementioned findings, although the internal alumina structure differs slightly, it can be concluded that both SGA03 and SGA90 exhibit a more networked filler structure compared to suspension-type thermal pastes.The dispersed silicone oil within glycerol prevents the collision of adjacent alumina particles, contributing to the distinctive internal structure observed in these emulsion-type thermal pastes (Figure 4; Figures S6 and S7, Supporting Information).Moreover, above 40 vol% alumina, unlike the suspension-type thermal pastes, SGA03 and SGA90 exhibited segregated and ordered filler structures, respectively (Figure 4).Owing to the formation of these structures, filler movement is significantly restricted; therefore, the yield stress increased more rapidly than the TC upon increasing the filler content, resulting in a sharp decrease in m values (Figure 3b; Table S3 in Supporting Information).As a result, as depicted in Figure 5, emulsion-type thermal pastes with segregated and/or ordered filler structures always exhibited a higher TC and yield stress than suspension-type thermal pastes, particularly at relatively high filler loadings.Please note that among the emulsiontype thermal pastes, SGA03 with a segregated filler structure induced by Pickering emulsion showed a higher TC and yield stress than SGA90.Hence, thermal pastes with 90 μm alumina particles were omitted from subsequent investigations.

Heat and Moisture Stability of the Thermal Pastes
As thermal pastes are exposed to continuous heating/cooling cycles, their thermal stability over time is essential.Commercial thermal pastes, which are composed of a simple mixture of a single-liquid-matrix and thermally conductive filler, are vulnerable to long-term operation at elevated temperatures. [17,19]When an excessive amount of heat is concentrated in the thermal paste, the matrix might flow out owing to reduced viscosity (known as "pump-out"), leading to phase separation.Furthermore, air occupying these void spaces within the thermal paste acts as an insulator and hinders heat transfer. [10,11]To evaluate the structural stability of the suspension-and emulsion-type thermal pastes against heat, their complex viscosities were measured.Figure 6a illustrates the experimental setup.Upon annealing, some thermal pastes, particularly the suspension-type thermal pastes, exhibited structural collapse leading to matrix leakage.Therefore, small-amplitude oscillatory shear (SAOS) measurements were conducted at room temperature on samples that had undergone prior thermal annealing.Prior to the measurements, each thermal paste was annealed for 1 h at 25, 50, 75, and 100 °C. Figure 6b-d exhibit the angular-frequency-dependent complex viscosities of several suspension-type thermal pastes (SA03_50 and GA03_50) and an emulsion-type thermal paste (SGA03_50), and Figure 6e shows their complex viscosities at an angular frequency of 0.1 rad s −1 as a function of preheating temperature.As shown in Figure 6b,c,e, the complex viscosity of SA03_50 increased as the annealing temperature increased, whereas the complex viscosity of GA03_50 sharply decreased as the annealing temperature increased.The former was attributed to a liquid matrix leakage, while the latter was caused by filler sedimentation.Both behaviors are a direct consequence of insufficient structural stability upon successive heating/cooling cycles.By contrast, owing to the segregated filler structure (Figure 4a,b), SGA03_50 retained its complex viscosity, even after preheating at temperatures up to 100 °C (Figure 6d,e).Furthermore, for the same reason, as the preheating temperature increased, both SA03_50 and GA03_50 exhibited significant variations in yield stress and TC upon heating and cooling, whereas SGA03_50 retained its original yield stress and TC (Figure 6f,g).
To further investigate the thermal stability of the thermal pastes in real time, Turbiscan analysis was performed.For this measurement, the samples were placed in cylindrical vials and exposed to near-infrared light (NIR) with a wavelength of 880 nm.Both transmitted (T) and/or backscattered (BS) NIR light passing through or scattering from the sample, respectively, was detected by a transmission and/or backscattering detector at every 40 μm of the sample height.Figure 6h-j and Figure S8a-c (Supporting Information) depict the time-dependent variations in T (ΔT) and the BS (ΔBS) signals from the initial value at each height for SA03_50, GA03_50, and SGA03_50, respectively, measured at 60 °C.This temperature was chosen because it is the maximum temperature that can be applied by the instrument.For the suspension-type thermal pastes (SA03_50 and GA03_50), the Figure 6.a) Schematic depicting the experimental procedure for assessing the thermal stability of the thermal pastes using a rheometer.All measurements were conducted at room temperature.Before testing, samples were subjected to heating at 25, 50, 75, and 100 °C for 1 h.b-d) Complex viscosity profiles of SA03_50, GA03_50, and SGA03_50 were measured using small-amplitude oscillatory shear (SAOS) mode at room temperature.Prior to measurement, samples were heated to 25, 50, 75, and 100 °C for 1 h.e-g) Complex viscosity at an angular frequency of 0.1 rad s −1 , yield stress, and TC of SA03_50, GA03_50, and SGA03_50 measured at room temperature.Before these measurements, all thermal pastes were thermally annealed at 50, 75, and 100 °C for 1 h.h-j) Transmittance profiles for SA03_50, GA03_50, and SGA03_50 as a function of annealing time at 60 °C using Turbiscan analysis.
ΔT and ΔBS signals at the upper height of the sample significantly increase and decrease over time, respectively (Figure 6h,i; Figure S8a,b, Supporting Information).This indicates that the fillers settled down, revealing a transparent matrix on top.On the other hand, for the emulsion-type thermal paste SGA03_50, no significant variations in ΔT and ΔBS were observed over time owing to its robust segregated filler structure (Figure 6j; Figure S8c, Supporting Information).In fact, as Figure S8d (Supporting Information) reveals, while SA03_50 and GA03_50 exhibited an obvious phase separation, SGA03_50 maintained its formation-even at 60 °C for 7 days.Furthermore, for quantitative analysis, the Turbiscan stability index (TSI), a specific parameter representing the physical stability of a suspension, was also calculated for each formulation; the calculation details can be found in the Experimental Section.Lower TSIs indicate a more uniform filler dispersion state, whereas higher TSIs indicate a heterogeneous filler dispersion state.As shown in Figure S8e (Supporting Information), the estimated TSI of SGA03_50 was less than 2, whereas those of SA03_50 and GA03_50 were significantly higher.Once again, this affirms that the emulsion-type thermal pastes were significantly more thermally stable than the suspension-type thermal pastes.
Along with heat exposure, electronics that operate in extreme environments can also be exposed to humid conditions. [53]herefore, thermal pastes should be stable not only against heat but also against moisture.Glycerol, one of the matrices used in this study, is a hygroscopic liquid owing to its polar hydroxyl functional groups, thus it is necessary to evaluate the pastes' stability against moisture.Hygroscopicity was evaluated by measuring the paste weight variations in a humid atmosphere over time; Figure 7a illustrates the experimental setup.The weight changes of GA03 and SGA03 exposed to an atmosphere of 95% relative humidity (RH) were recorded over time at room temperature.The hygroscopicity of GA03 and SGA03 was estimated using Equation (3).
where W t and W 0 are the instantaneous and initial weights of the thermal pastes, respectively.As shown in Figure 7b,c, the hygroscopicity and normalized mass with respect to the initial mass for both GA03 and SGA03 increased over time, regardless of alumina content.Moreover, the water uptake of GA03 and SGA03 decreased as their alumina loadings increased, indicating that alumina clearly hindered water uptake.Owing to the high hygroscopicity of glycerol, the mass of GA03_50 increased by ≈20% compared with its initial value.By contrast, the masses of the SGA samples only increased by 7% compared with their initial masses-even SGA03_10, which had the lowest filler load-ing.Furthermore, only a 1% mass increase was observed for SGA03_50.This is because its segregated filler structure with a tortuous pathway effectively hindered moisture penetration to the matrix.Owing to these factors, we ascertained that SGA03, with its segregated filler structure, exhibited remarkable stability against both heat and moisture, making it suitable for use as a form-factor-free thermal paste with high TC.

Recyclable Pickering Emulsion with High TC as a Form-Factor-Free Thermal Paste
Among the thermal pastes evaluated in this study, SGA03, characterized by a Pickering emulsion with a well-segregated filler structure, exhibited a higher TC as well as a higher stability against both heat and moisture than the other thermal pastes.Therefore, in order to use SGA03 as a thermal paste in practical applications, it should be extrudable through a syringe with a continuous flow while retaining its structural stability.As shown in Figure 8a, SGA03_50 was successfully extruded through a syringe without undergoing any cut-off.Despite its high yield stress and complex viscosity (Figure 2a; Figure S1, Supporting Information), SGA03_50 could be extruded through simple compression, enabling its use as a thermal paste.Moreover, to evaluate the thermal stability of SGA03_50 after its application, the extruded paste was thermally annealed at 100 °C for 7 days (Figure 8b).As shown in Figure 8c, extruded SGA03_50 molded into the shapes of the letters "P," "N," and "U" on the aluminum plate retained their shape even after 7 days of thermal annealing at 100 °C, without undergoing any matrix leakage and/or phase separation.This result is consistent with previous rheological and Turbiscan results (Figure 6), highlighting the high thermal stability of SGA03_50.It also indicates that the internal filler structure was preserved, even after extrusion through a syringe.As can be seen from Figure S9 and Movie S1 (Supporting Information), SGA03_30 was successfully 3D printed without clogging, resulting in the formation of a stacked-ring structure (SGA03_30 with a viscosity of ≈10 5 Pa s was used, following the manufacturer's suggested viscosity range for the 3D printer).Coupled with syringe extrusion and 3D printing, SGA03_50 exhibited a putty-like formulation, highlighting the thermal stability of SGA03_50 after syringe application onto an aluminum plate.c) Photographs of the extruded SGA03_50 forming the letters "P," "N," and "U" on an aluminum plate before and after exposure to 100 °C for 7 days.d) Schematic illustrating the molding process and the evaluation of thermal stability for SGA03_50 across multiple heating/cooling cycles.e) Photographs displaying the molded SGA03_50 in three distinct shapes: heart, leaf, and cat.f) Surface temperature profile of the molded SGA03_50.Each shape underwent remolding after two successive heating/cooling cycles, amounting to a total of six cycles.The solid and dashed lines indicate the heating and cooling process, respectively.The colors pink, green, and gray correspond to the heart, leaf, and cat shapes used during the heating/cooling cycles.g) IR thermal images of the thermal paste, which aligned with the aforementioned six consecutive heating and cooling cycles.
which could be kneaded without leaving any residue (Figure S10, Movie S2 in Supporting Information).Therefore, SGA03_50 can be easily processed into various shapes through a simple molding process using a silicone mold.
To demonstrate the long-term thermal stability and heatdissipation performance of SGA03_50 after molding, molded SGA03_50 was placed on a heating plate and then subjected to successive heating/cooling cycles by alternatively switching the heating power.To further highlight its reprocessability, the initially molded configuration (a heart shape) of SGA03_50 was subsequently remolded into two distinct shapes (a leaf and then a cat), following each two successive heating/cooling cycles (Figure 8d,e).Note that the putty-like formulation of SGA03_50 was retained even after being subjected to multiple heating/cooling cycles, enabling it to be remolded into different shapes (Figure 8e).The time-dependent surface-temperature profiles of the molded SGA03_50 during the heating/cooling process were recorded using a calibrated IR thermal imaging camera (Figure 8f,g); the details are provided in the Experimental Section.The surface-temperature variation rate of the molded SGA03_50 did not change significantly over a total of six sequential heating/cooling cycles (Figure 8f,g).
The low recyclability of composites is one important challenge that needs to be addressed.So far, only energy-intensive recycling processes have been applied, which are required to overcome the strong bonding between the matrix and fillers and come with a high entropy penalty. [54]However, the thermal pastes used in our study possessed flowability and relatively weak matrix-filler interactions.Hence, the matrix and filler can be recycled via facile centrifugation.Figure S11a (Supporting Information) illustrates the recycling process for recovering alumina from SGA03_50.To ensure sufficient flowability, n-hexane and deionized (DI) water, which can dissolve silicone oil and glycerol, respectively, were added to SGA03_50, followed by vigorous mixing using a planetary centrifugal mixer.After the dilution process, centrifugation was performed three times to separate the matrix (a mixture of silicone oil and glycerol) and alumina based on their density difference.The supernatant was then removed, and vacuum filtration/drying was carried out to remove the remaining liquid matrix from the precipitated alumina.After collecting the alumina, the thermal properties of the recovered alumina and the stock alumina were compared using TGA analysis.As shown in Figure S11b (Supporting Information), no significant differences in weight loss were observed upon heating to 800 °C compared to the stock alumina.This outcome confirms that alumina was successfully recovered via centrifugation without any residue and/or quality deterioration.

Conclusion
In summary, we reported a new strategy for the preparation of highly thermally conductive, stable thermal pastes based on Pickering emulsions with segregated filler structures.Two immiscible liquids, silicone oil, and glycerol, were used along with a costeffective, thermally conductive filler, alumina.The suspensiontype thermal pastes (or a single-liquid-matrix and randomly dispersed fillers) showed low stability against heat and moisture along with relatively low TC, even at high filler loadings.However, the Pickering-emulsion-based thermal pastes exhibited a significant enhancement in both TC and stability against heat and moisture, owing to the segregated filler structure.Moreover, the high applicability of the emulsion-type thermal pastes enabled syringe extrusion, 3D printing, multiple reprocessing/molding, and eco-friendly recycling.As part of the emulsion preparation through simple mixing, the segregated structure could be simultaneosuly formed.Therefore, this approach can be readily adopted by relevant industries to improve myriad composite functionalities originating from the filler-network structure, such as gas-barrier properties, electrical conductivity, and mechanical properties.Although alumina was selected here as a representative cost-effective ceramic filler, it is important to acknowledge the possibility of utilizing other high-TC fillers, including hexagonal boron nitride, MXenes, and silver.Accordingly, this Pickering emulsion approach can represent a general strategy for enhancing the TC, stability, and applicability of nextgeneration, high-performance TIMs.
Preparation of Suspension-Type Thermal Pastes: Silicone oil and/or glycerol were first weighed into a 24 mL plastic container.Then, 10, 20, 30, 40, and 50 vol% alumina of varying sizes was added with respect to the total paste volume.The paste was then vigorously mixed for 1 min at 2000 rpm using a planetary centrifugal mixer or a Thinky mixer (AR-100, Thinky, Japan).To prevent the formation of voids or bubbles in the paste, an additional defoaming and mixing procedure was performed for 1 min at 2200 rpm and 1 min at 2000 rpm.The resulting pastes maintained their flowability upon tilting, even at the maximum filler loading used in this study (50 vol% alumina).
Preparation of Emulsion-Type Thermal Pastes: The emulsion-type thermal pastes were prepared by first vigorously mixing silicone oil and glycerol in a volume ratio of 1:1 for 3 min at 2000 rpm using a Thinky mixer.Then, 10, 20, 30, 40, and 50 vol% alumina of varying sizes with respect to the total paste volume were added to the mixture of the two immiscible liquids, i.e., silicone oil and glycerol.These components were then mixed together using the same mixing sequence used for the suspensiontype thermal pastes.At all filler loadings, the final stabilized emulsion-type thermal pastes did not flow, even upon tilting.
TC Measurement of Thermal Pastes: The TC of the thermal pastes was measured using a transient plane source method (TPS-2500S, Hot-disk AB, Sweden) in isotropic mode at room temperature.Owing to the fluid nature of the thermal pastes, they were wrapped in plastic wrap and dispensed into two plastic caps in order to preserve the formulations.The depth and the diameter of each plastic cap were 10 mm and 20 mm, respectively.During the measurement, two identical plastic caps filled with thermal paste were sandwiched between the Kapton sensor sizes with a maximum radius of 3.2 mm (5465 probes).The output heating power applied to the sensor was ≈0.2 W, and the measurement time was 10 s, resulting in a change of temperature and contact resistance between the sensor and thermal pastes.Then, the thermal diffusivity and TC were calculated from the mathematical model.Three tests were performed for each thermal paste, and the average TC was reported.
Rheological Properties of Thermal Pastes: Rheological properties such as complex viscosity and yield stress were measured in small-amplitude oscillatory shear (SAOS) mode using a rotational rheometer (HR-20, TA Instrument, USA) equipped with 40-millimeter-diameter parallel plates.Approximately 2 mL of each sample was subjected to oscillatory shear of varying angular frequency, and the dynamic moduli and complex viscosity change as a function of angular frequency were observed.The thermal pastes are typical Herschel-Bulkley fluids, exhibiting yield stress at low angular frequencies and power-law model-like behavior at medium and high angular frequencies.The yield stress of each thermal paste was realized by fitting the complex viscosity data to the Herschel-Bulkley model, as shown in Equation ( 4). [46] where *(),  0 ,  y , , k, and n are the complex viscosity as a function of angular frequency, the zero-shear viscosity, the yield stress, the angular frequency, the flow coefficient, and the Herschel-Bulkley index, respectively.The complex viscosity as a function of angular frequency and the predicted complex viscosity obtained from the Herschel-Bulkley model are shown in Figure S1 (Supporting Information).
Internal Filler Structure Within the Thermal Pastes: The internal filler structure of each thermal paste was observed using a confocal laser scanning microscope (LSM800, Zeiss, Germany).To demonstrate the difference between the internal morphology of the suspension-type and emulsion-type thermal pastes, GA and SGA were observed.For GA, Nile blue was added, followed by mixing with a Thinky mixer at 2000 rpm for 1 min.For SGA, to distinguish silicone oil (oil phase) and glycerol (water phase), Nile red and Nile blue were added, respectively, with subsequent mixing using the Thinky mixer (2000 rpm, 1 min).Note that the amount of Nile red and Nile blue within the thermal pastes is 0.1% with respect to the total sample volume.Afterward, the stained thermal pastes were uniformly deposited onto glass slides, and glass coverslips were placed over the samples.The thermal pastes were exposed to various light sources, including white light and laser beams (488 nm and 561 nm wavelengths).Note that the excitation wavelengths of Nile red and Nile blue are 488 nm and 561 nm, respectively.In addition, the thermal pastes containing alumina with average diameters of 3 and 90 μm were observed under 20× and 10× magnification, respectively.For each sample, images were recorded under white light (bright field) and laser irradiation at 488 nm, 561 nm, and both 488 nm and 561 nm.
Tilting Test: A tilting test was performed for SA03_50, GA03_50, and SGA03_50 to compare the structural stability of the suspension-type and emulsion-type thermal pastes.Samples were applied to a commercial paper box, and the box was tilted in the vertical direction (angle of 90°) and observed for 7 days.Unlike the suspension-type thermal pastes, SGA03_50 did not flow upon tilting, even after 7 days.
Heat Stability Test: To investigate the heat stability of the thermal pastes, they were thermally annealed at 50, 75, and 100 °C for 1 h on a Peltier plate in a rheometer (HR-20, TA Instrument, USA) and then cooled to room temperature.After sufficient cooling, the rheological properties were measured using the process described above.Furthermore, to observe the variation of TC upon heating, the thermal pastes prepared for the TC measurements were annealed on a hot plate at 50, 75, and 100 °C for 1 h.The TC was then measured using the same procedure as described above.
Turbiscan Analysis: The structural stability of the thermal pastes against heat was measured using a dispersion stability analyzer (Turbiscan Lab Expert, Formulaction, France).Approximately 20 mL of each thermal paste was filled into a cylindrical vial, retaining a flat meniscus.After setting the instrument's oven temperature to 60 °C, the vial was inserted.Then, the detector head with a pulsed NIR light source (880 nm) and two synchronous detectors (transmission and backscattering) were moved in the height direction of the vial.Transmission and backscattering signals were collected every 40 μm with an interval of 40 min for 7 days.The TSI, a specific parameter indicating the physical stability of a suspension, was automatically calculated by the Turbiscan software.TSI values can be derived from the variations in the intensity of transmitted and backscattered light with respect to the original intensity (Equation 5).TSI = √ ∑ n i=1 (X i − X BS ) 2 n − 1 (5)   where X i , X BS , and n is the backscattering signal intensity at the time i, the average value of X i , and scan number, respectively.Moisture Stability Test: Moisture stability was realized by observing the weight change of the thermal pastes under humid conditions.Humidity was measured using a commercial hygrometer.The weights of the thermal pastes were measured using an analytical balance (AUW220D, Shimadzu, Japan).During the measurement, four opened vials containing DI water were placed at each corner of the balance.After the steady-state RH in the balance reached 95%, the change in mass of the thermal pastes was recorded every 10 min at room temperature.
3D Printing of Thermal Pastes: A thermal paste formulated as an "inktype" filament was 3D printed using a Ultimaker 2+ printer (Ultimaker, Netherlands) equipped with a Discov3ry paste extruder.The sample paste was charged in a 50 mL syringe and extruded on a target substrate at 1.3 mm −1 s through a nozzle (diameter: 610 μm) to construct the 3D structure designed using Cura 4.1.0software.
Cyclic Heating/Cooling Test of Molded Thermal Pastes: SGA03_50 was molded using commercial silicone molds of different shapes (heart-, leaf-, and cat-shaped).First, SGA03_50 was poured into the heart-shaped silicone mold.After SGA03_50 assumed the heart shape, the molded SGA03_50 was de-molded by applying gentle pressure to the silicone mold.Then, heart-shaped SGA03_50 was placed on the commercial heating plate and heated from room temperature to 100 °C for 10 min.After heating, SGA03_50 was cooled to room temperature for 40 min by turning off the heating plate.This heating/cooling process was performed one more time.After the heart-shaped SGA03_50 underwent two heating/cooling cycles, it was removed and remolded into a leaf shape in the same manner as described previously.Then, two heating/cooling cycles were carried out in the same manner.Finally, the leaf-shaped SGA03_50 was removed again and remolded into a cat shape, followed by two more heating/cooling cycles.Thus, six cyclic tests were performed, i.e., two cyclic heating/cooling tests for each of the three differently shaped SGA03_50 samples.Throughout each cycle, thermal images were captured by an IR thermal imaging camera (E5-XT, FLIR, USA) to evaluate the heat-dissipation performance of the molded SGA03_50 upon heating.
Filler Recycling: SGA03_50 was diluted by adding 20 mL of n-hexane and DI water.The mixture was then vigorously mixed at 2000 rpm for 3 min using a Thinky mixer (AR-100, Thinky, Japan).Afterward, the diluted mixture was centrifuged at 4000 rpm for 10 min using a centrifuge (1248R, LABOGENE, Denmark); this process was repeated three times.The phase-separated matrix was then removed from the mixture.To completely remove the matrix within the precipitated filler, the collected filler was washed with n-hexane and DI water several times under vacuum filtration.Finally, the washed filler was dried at 100 °C in a vacuum oven (C-DVD1, Changshin Science, Republic of Korea) for 1 day.The quality of the recovered alumina and the stock alumina was measured using thermal gravimetric analysis (TGA; Q50, TA Instrument, USA) at a ramp rate of 10 °C min −1 .Approximately 10 mg of the sample was placed on a platinum pan and heated from 30 to 800 °C under a nitrogen atmosphere.

Figure 1 .
Figure 1.a) Schematic depicting the preparation process for suspension-and emulsion-type thermal pastes.b,c) Isotropic TC as a function of alumina loading for thermal pastes.The corresponding results for the thermal paste containing alumina with an average diameter of 3 μm (SA03, GA03, and SGA03) and 90 μm (SA90, GA90, and SGA90) are shown in b) and c), respectively.d-f) Experimental and theoretical TC for SA, GA, and SGA, respectively.The square and triangle symbols represent thermal pastes containing alumina microparticles with average diameters of 3 and 90 μm, respectively.The theoretical TC as a function of alumina loading was predicted using the Bruggeman model (dashed line), as shown in Equation(1).In this prediction, the filler TC was held constant at 30 W m −1 K, the TC value of alumina reported in the literature.[13]

Figure 2 .
Figure 2. a,b) Yield stress as a function of alumina loading for thermal pastes containing alumina with average diameters of 3 μm (SA03, GA03, and SGA03) and 90 μm (SA90, GA90, and SGA90), respectively.The blank symbols indicate filler-free samples with Newtonian flow behavior, where the yield stress is ≈0.c) Photographs depicting SA03 and SGA03 at alumina loadings ranging from 10 to 50 vol%.These images were taken after the thermal pastes rested for 7 days at room temperature.SA03 underwent phase separation for all filler loadings, while SGA03 maintained its paste-like formulation for all filler loadings.d) Photographs capturing SA03_50, GA03_50, and SGA03_50 during the tilting test.Upon tilting motion at an angle of 90°, SA and GA exhibited flowability even at the highest filler loading (50 vol%), whereas SGA03_50 did not flow for 7 days at room temperature upon tilting.

Figure 2a ,
Figure 2a,b as a function of filler loading (the corresponding values are summarized in TableS2, Supporting Information).Note that pure silicone oil and glycerol showed Newtonian behavior, where the viscosity is a function of angular frequency, resulting in virtually no yield stress.On the contrary, the mixture of silicone oil and glycerol (unstable emulsion form) exhibited a yielding behavior owing to the restricted translational and rotational motion of the dispersed droplets (Figure2a,b; FigureS1, Supporting Information).As depicted in Figure2a,b, the yield stress increased as the filler loading was increased for both the suspension-and emulsion-type thermal pastes, owing to the matrix-filler and/or filler-filler interactions.The emulsion-type thermal pastes, however, always exhibited higher yield stresses than the suspensiontype thermal pastes.For both SGA03 and SGA90, remarkable increments in yield stress were noted within the 0-10 vol% range of alumina loading.Such significant enhancements in yield stress were not detected in the suspension-type thermal pastes.These observations collectively suggest the establishment of a more robust filler-network structure within the SGA samples, even at alumina loadings below 10 vol%.The influence of greater yield stress in emulsion-type thermal pastes can be further substantiated through a comparison of particle stability against gravitational sedimentation.Figure2cshows photographs of SA03 and SGA03 after 7 days of resting at room temperature.FigureS2(Supporting Information) shows the other suspension-and emulsion-type thermal pastes with

Figure 3 .
Figure 3. a) Experimental TC of the thermal pastes as a function of the corresponding yield stress.The Prasher model (Equation 2) predictions are also provided on the same graph.The solid lines and dashed lines represent the fitting outcomes of the Prasher model, applied before and after reaching 40 vol% alumina loading.b) Predicted m values of the thermal pastes using the Prasher model.Given the distinct behavior of the thermal pastes with alumina loadings of less than and greater than 40 vol%, the m values were independently estimated by fitting the experimental TC and the corresponding yield stress for each scenario.
ning microscope.Confocal micrographs for SGA03 and SGA90 are displayed in Figure 4, while those for GA03, and GA90 are shown in Figures S4, and S5 (Supporting Information), respectively.More detailed confocal micrographs of SGA03 and SGA90, captured under light sources with different wavelength, are shown in Figures S6 and S7 (Supporting Information), respectively.The numbers appearing after the underscore in each sample label indicate the alumina content in volume percentage.

Figure 4 .
Figure 4. a,c) Confocal micrographs and b,d) corresponding schematic representations of SGA03 and SGA90 with filler loadings ranging from 10 to 50 vol%.A confocal laser scanning microscope equipped with a 20× and 10× magnification lens was employed for SGA03 and SGA90, respectively.Each column, from left to right, displays SGA03 and SGA90 containing 10, 20, 30, 40, and 50 vol% alumina.In the confocal micrographs, Nile-red-stained silicone oil appears green, while Nile-blue-stained glycerol appears red.The scale bars in a) and c) represent 50 and 100 μm, respectively.In the schematic illustrations, yellow, blue, and gray correspond to silicone oil, glycerol, and alumina, respectively.

Figure 5 .
Figure 5. Schematic illustrating the internal filler structure and the estimated heat-transfer pathway within suspension-type thermal pastes with randomly dispersed filler and emulsion-type thermal pastes with a segregated filler structure.In suspension-type thermal pastes, a discontinuous heat-transfer pathway is observed, while emulsion-type thermal pastes possess a continuous and robust heat-transfer pathway.

Figure 7 .
Figure 7. a) Photograph of the experimental setup for measuring weight fluctuations of the thermal pastes in a humid environment.b) Hygroscopicity,and c) normalized mass of GA03 and SGA03 with respect to the initial weight.The alumina filler loadings ranged from 10 to 50 vol%.All measurements were conducted at room temperature and 95% relative humidity (RH).

Figure 8 .
Figure8.a) Photograph depicting the extrusion of SGA03_50 through a syringe using gentle pressure.b) Schematic outlining the experimental setup, highlighting the thermal stability of SGA03_50 after syringe application onto an aluminum plate.c) Photographs of the extruded SGA03_50 forming the letters "P," "N," and "U" on an aluminum plate before and after exposure to 100 °C for 7 days.d) Schematic illustrating the molding process and the evaluation of thermal stability for SGA03_50 across multiple heating/cooling cycles.e) Photographs displaying the molded SGA03_50 in three distinct shapes: heart, leaf, and cat.f) Surface temperature profile of the molded SGA03_50.Each shape underwent remolding after two successive heating/cooling cycles, amounting to a total of six cycles.The solid and dashed lines indicate the heating and cooling process, respectively.The colors pink, green, and gray correspond to the heart, leaf, and cat shapes used during the heating/cooling cycles.g) IR thermal images of the thermal paste, which aligned with the aforementioned six consecutive heating and cooling cycles.