1 Introduction

Although polymers have a wide range of applications in electronic devices, including power electronics, electric motors, and generators due to their lightweight, corrosion resistance, and ease of processing [1,2,3,4], their low thermal conductivities (~ 0.2 W m−1 K−1) severely limit their heat conduction and dissipation in electronic devices. Therefore, the improvements in thermal conductivity of polymer materials, especially their through-plane thermal conductivity, are highly crucial for their application as thermal interface materials (TIMs) between heaters and heat sinks [5]. To achieve this goal, metallic, ceramic, and carbon-based thermally conductive fillers, such as silver nanoparticles [42], where λ is the laser wavelength, and ID/IG is the integrated intensity ratio of D band to G band. Thermal conductivities were calculated by k = α × ρ×Cp, where α is the thermal diffusivity, Cp is the specific heat capacity, and ρ is the density. Thermal diffusivities of GEs were measured on a Netzsch LFA467 light flash apparatus from 30 to 80 °C. Specific heat capacities of GEs from 30 to 80 °C were obtained using a TA Q20 differential scanning calorimeter (DSC) at a scanning rate of 10 °C min−1. Densities of GEs were measured by an Mettler Toledo electronic balance with a density determination kit 33,360. Thermogravimetric analysis (TGA) curves were obtained with a TA Q50 thermogravimetric analyzer at a heating rate of ~ 10 °C min−1 in a nitrogen atmosphere. The volume fraction was calculated by: vol% = wt% × (ρcom/ρ*), where wt% is the mass fraction, ρcom is the density of a composite, and ρ* is the true density of graphene (2.25 g cm−3). The heat transfer performances of GEs were recorded by a FLIR E40 infrared camera. High-resolution transmission electron microscopy (HRTEM) images were obtained using a JEM-2100Plus microscope at an operation voltage of ~ 200 kV. Three-point bending tests (span 7.75 mm) were performed on a SUNS UTM4103 tester with a loading rate of 0.05 mm min−1. The loading direction is perpendicular to the lamellar surface of the LSGA. For single-edge notched beam specimens, the epoxy and GEs were cut and polished to beams with dimension of 12 × 2.5 × 2.5 mm3, and the beams were then notched by a diamond blade. The notch was sharpened by a razor blade perpendicular to the lamellar direction. The notch was about half of the thickness of specimens. The initial fracture toughness (KIC) was calculated by Eq. (1) [43]:

$$K_{IC} = \frac{{P_{IC} S}}{{BW^{3/2} }}f\left( {\frac{a}{W}} \right),\,\,f\left( {\frac{a}{W}} \right) = \frac{{3\left( {\frac{a}{W}} \right)^{1/2} \left[ {1.99 - \frac{a}{W}\left( {1 - \frac{a}{W}} \right)\left( {2.15 - 3.93\frac{a}{W} + 2.7(a/W)^{2} } \right)} \right]}}{{2\left( {1 + 2\frac{a}{W}} \right)\left( {1 - \frac{a}{W}} \right)^{3/2} }}$$
(1)

where PIC is the maximum load before crack initiation, B is the width, S is the span, W is the thickness, and a is the notch depth of the specimens. The maximum fracture toughness (KJ) was calculated by Eq. (2) [44]:

$$K_{J} = \sqrt {\frac{{EJ_{PI} }}{{1 - \nu^{2} }} + K_{IC}^{2} } ,\,\,J_{PI} = \frac{{2A_{PI} }}{{B\left( {W - a} \right)}}.$$
(2)

where API is the area under the force–displacement curve, E is the Young’s modulus, and ν is the Poisson’s ratio. The crack extension (Δa) is calculated by Eq. (3) [44]:

$$a_{n} = a_{n - 1} + \frac{{W - a_{n - 1} }}{2}\frac{{C_{n} - C_{n - 1} }}{{C_{n} }},\,\,C_{n} = \frac{{u_{n} }}{{f_{n} }},\,\,{{\Delta a}} = a_{n} - a$$
(3)

where \(a_{n}\) is the crack length, \(u_{n}\) is the displacement, and \(f_{n}\) is the force at each point after crack initiation.

3 Results and Discussion

3.1 Morphologies and Microstructures of Lamellar-Structured Graphene Aerogels

Figure 1a illustrates the fabrication of LSGAs and their epoxy composites. During the bidirectional freezing of the PAAS/GO suspension, ice crystals nucleate and grow to be parallel lamellae because of the temperature gradients in both horizontal and vertical directions, expelling the PAAS and GO components from the ice crystal lamellae to replicate the lamellar morphology [45, 46]. By freeze-drying to remove the ice crystals by their subliming, the resultant PAAS/GO hybrid aerogels are thermally annealed at 300 °C, during which the PAAS monomers are polymerized to PI, while the GO component is partially reduced to RGO. The bidirectionally orientated PI/RGO hybrid aerogel is less thermally conductive because of the less conductive PI and the poor conductivity of RGO. Therefore, the resultant PI/RGO hybrid aerogel is graphitized at 2800 °C to carbonize and even graphitize the thermally insulating PI macromolecules and to convert RGO to high-quality graphene by removing its residual oxygen-containing groups and healing its lattice defects.

As expected, the PAAS/GO hybrid aerogels show lamellar and porous structures, and the spacing between adjacent lamellae varies in the range of 20-40 μm (Fig. 1b–f). It is also seen that the initial dosage of GO greatly affects the lamellar structure of the hybrid aerogels, especially when the dosage of GO exceeds 50 wt%. As shown in Fig. S2, P6G4 and P5G5 have long-range lamellar structures (Fig. S2a, b). Further increasing the GO dosage tends to form disordered structures (Fig. S2c, d). Particularly, a neat GO aerogel derived from a concentrated GO suspension (40 mg cm−3) in the absence of PAAS presents a completely disordered structure (Fig. S2e, j). This is not only because the oxygen-containing groups on the GO sheets are willing to be adsorbed to the surface of the ice crystal, and hence, causing a curved ice crystal during freezing process, the high viscosity of GO suspension at such a high concentration also hinders the unidirectional growth of ice crystal [47]. Fortunately, after the imidization and graphitization treatments, the anisotropic LSGAs still maintain their lamellar structures (Fig. 1g–k). The bidirectional orientation extents of the lamellae are affected by initial PAA/GO mass ratios. P9G1-2800 and P8G2-2800 show a waved multi-arch morphology. With increasing the GO dosage, the lamellae change from crumpled to relative flat. This is because the volume decrease in PI component during the graphitization process is larger than that of GO sheets, giving rise to uneven stress distribution in the lamellae, especially at high PI contents [39]. Consequently, a lamellar-structured high-quality graphene aerogel is thus fabricated (Fig. S3).

For graphene aerogels used as thermally conductive fillers, their apparent density determines the graphene content in composites, which affects the thermal conductivities of resultant composites. Here, because the volume decrease in GO is less than that of PI during their graphitization process, the apparent density of LSGAs decreases from ~ 40.9 to ~ 16.6 mg cm−3 as the GO content increases from 10 to 50 wt% (Fig. 2a). In addition to the apparent density of the LSGAs, the graphene quality is also crucial for efficient heat conduction along the aerogel skeleton walls. XRD is used to characterize the graphene quality by probing the amount and orientation of graphitic carbon layers and the curvature of individual sheets [50]. Different from Raman spectrum at single points, Raman map** reflects the distribution of lattice defects more accurately in a certain area. As shown in Fig. 3a–e, different colors represent different ID/IG ratios. The blue, green, and red colors correspond to their ID/IG values of 0, ~ 0.2, and ~ 0.4, respectively. Compared to PAA-2800 with an average ID/IG of ~ 0.189 (Fig. S4a), the graphitization extent of P9G1-2800 is greatly improved with a low average ID/IG of ~ 0.087, although it still presents large green and glaucous areas. Interestingly, as the increase in the GO dosage, the blue area becomes larger, while the green area decreases gradually. The average ID/IG decreases to ~ 0.028 for P5G5-2800, very close to that of GO-2800 (~ 0.026) (Fig. S4b). Meanwhile, as shown in Fig. 3f, the decreased intensity of ID/IG manifests the stepwise healing of sp2 domains from ~ 87.4 nm of PAA-2800 to ~ 188.9 nm of P9G1-2800, and then to ~ 583.4 nm of P5G5-2800 [42]. Apparently, these Raman map**s indicate that graphitization converts PI to graphitized carbon and reduce GO to high-quality graphene. It is also seen that GO plays an inducing role in the conversion of PI to graphitized carbon because the large sheet of GO could promote the orientation of PI macromolecules and thus improve the graphitization extent of LSGAs [51, 52].

Fig. 3
figure 3

Raman map**s of a G9P1-2800, b G8P2-2800, c G7P3-2800, d G6P4-2800, and e G5P5-2800. f Plots of average ID/IG value and crystal size of LSGAs as a function of GO content in PAAS/GO suspensions. g TEM and h HRTEM images of G6P4-2800

Full size image

In addition to the low lattice defects, and large graphene crystal size, the high quality of the vertically aligned skeleton walls is also crucial for thermal conduction of LSGAs. HRTEM is adopted to observe the skeleton wall of P6G4-2800 (Fig. 3g, h). The lamella thickness is ~ 29.8 nm (Fig. 3g), and the P6G4-2800 possesses vertically aligned and closely stacked graphitic lamellae, which is similar to a highly thermally conductive graphitic film [36]. The closely stacked skeleton walls of P6G4-2800 would benefit the decrease in contact thermal resistances. It is believed the PAAS joints and fills the gaps between the GO sheets and both of them are expelled to form vertically aligned and closely stacked lamellae by generated ice crystals during the bidirectional freezing, and the closely stacked lamellae are converted to orderly stacked graphitic layers after the imidization and graphitization treatments (Fig. 3g) [53]. Moreover, due to the ultrathin lamellae and the highly porous structure of the PAAS/GO hybrid aerogel, the gases released during the imidization and graphitization process may escape easily, which would not cause the gases accumulation and the froth of the closely stacked lamellae [56]. Table S2 lists through-plane thermal conductivities of the LSGA/epoxy composites, the average ID/IG values of LSGAs, and the graphene contents calculated on the basis of the TGA curves (Fig. S9). As shown in Fig. 4a and Table S2, both the quality of LSGAs and the graphene content affect the ultimate thermal conductivity of the epoxy composites. To eliminate the effect of graphene content and highlight the role of the quality of LSGAs, the efficiency of thermal conductivity enhancement (η) is regarded as specific TCE and calculated by Eq. (4):

$$\eta = \left[ {\left( {K - K_{m} } \right)/\left( {100 \, VK_{m} } \right)} \right] \times 100\%$$
(4)

where Km (W m−1 K−1) and K (W m−1 K−1) are thermal conductivities of epoxy and LSGA/epoxy composites, respectively, and V (vol%) is the volume fraction of fillers. As shown in Fig. 4b, GE5 has the highest specific TCE of ~ 5054%, resulting from the highly efficient thermal conduction path of P5G5-2800. Compared to other composites shown in Fig. 4a, GE4 exhibits the highest thermal conductivity of ~ 6.51 W m−1 K−1 with a relatively high specific TCE of ~ 4750% at the graphene content of ~ 1.23 wt%, because GE4 owns highly efficient thermal conduction paths as well as relatively high filler content.

Fig. 4
figure 4

a Thermal conductivities along Z-direction, and b specific TCEs of graphene/epoxy composites. The data in a are graphene contents in their epoxy composites. c Comparison of thermal conductivities of GE4, GE4-30%, GE4-50%, and GE4-70% in three directions. d Plots of thermal conductivity of the composites in three directions as a function of graphene content. e Thermal conductivities of GE4-70% in three directions at different temperatures. f Comparison of thermal conductivity of GE4-70% in Z-direction with those reported in the literature

Full size image

To further improve thermal conductivity of the epoxy composites while maintaining their mechanical properties, the graphene content can be increased by slowly compressing the LSGAs perpendicular to the lamellar direction before the epoxy resin is thermally cured. As shown in Fig. 4c, d, the compression extents of P6G4-2800 can be 30%, 50%, and 70%, and the graphene contents in the epoxy composites increase from ~ 1.23 wt% of GE4 to ~ 4.28 wt% of GE4-70%. Meanwhile, the thermal conductivities in Z- and X-directions of the composites enhance with increasing the compression extents. For example, the thermal conductivity of GE4 along Z-direction is ~ 6.51 W m−1 K−1, while that of GE4-70% increases to ~ 20.0 W m−1 K−1, because the lamellar sheets become more compact, while the nacre-like structure of the composites is well retained (Fig. S5d-f) [57]. Note that the thermal conductivities do not change significantly in Y-direction with increasing the graphene content. As shown in Fig. 4d, the thermal conductivity of GE4-70% in Z-direction is ~ 20.0 W m−1 K−1, while that of GE4-70% in Y-direction is only ~ 1.22 W m−1 K−1, which is ascribed to the high interface thermal resistances between graphene and polymer matrix in the conduction path of the Y-direction.

The influence of temperature on thermal conductivity of the thermosetting epoxy composites is also evaluated (Fig. 4e). When the temperature increases from 30 to 80 °C, the thermal conductivity of GE4-70% decreases slightly in three directions. Nevertheless, the through-plane thermal conductivity (Z-direction) of GE4-70% at 80 °C is still as high as ~ 19.2 W m−1 K−1. As shown in Fig. 4f, the GE4-70% with ~ 4.28 wt% (2.30 vol%) of graphene exhibits a high through-plane thermal conductivity of ~ 20.0 W m−1 K−1, much higher than those reported in the literature at similar graphene contents [14, 19, 24,25,26, 28,29,30, 32, 33, 41, 58,59,60,61]. Moreover, GE4-70% presents a record-high specific TCE of ~ 4310% at such a high thermal conductivity among all kinds of fillers, which is even higher than those of film-type composites that usually have high in-plane thermal conductivities (Table S3) [24, 27, 59, 62]. Due to the superior elasticity of the lamellar structure, the graphene content could be tuned by compression along Z-direction without damaging the closely stacked graphene lamellae, and the enhancement efficiency only decreases a little when the graphene content increases from ~ 1.23 to ~ 4.28 wt%.

To illustrate the difference in thermal conductivities intuitively, the epoxy and its composites with a dimension of 10 × 10 × 10 mm3 are placed on the same hot stage at 75 °C, and an infrared camera is adopted to record in situ the side temperature variation of the samples (Fig. 5a, Movie S1). Figure 5a shows the infrared images after heating for 1, 30, 60, 100, and 160 s. The side temperature of GE4-70% along Z-direction (GE4-70%-Z) increases much faster than others, which should be attributed to its high through-plane thermal conductivity resulted from the high-quality vertically orientated graphene and the high graphene content of 4.28 wt%. The GE4-70%-Z is highly promising as TIM because of its exceptional high through-plane thermal conductivity. As shown in Fig. 5c, commercial silicone rubber with a thermal conductivity of ~ 6 W m−1 K−1 and GE4-70%-Z are inserted between a 10 W LED chip and a Cu heat disk. The thickness of the TIM is ~ 2 mm, and the LED chip/TIM/Cu heat disk interfaces are glued by a thermally conductive silicone grease. The surface temperatures of the LED chips are recorded with an infrared camera upon lightening (Fig. 5b). The series of infrared images reveal that the temperature increases sharply with the silicone rubber as TIM as compared to GE4-70%-Z. With the commercial silicone rubber as the TIM, the final surface temperature is up to ~ 84.5 °C, whereas the temperature is only ~ 71.3 °C in the presence of GE4-70%-Z (Fig. 5d) because of the excellent heat dissipation ability of GE4-70% along the Z-direction.

Fig. 5
figure 5

a Infrared images of epoxy and its composites on the same hot stage at 75 °C, showing that GE4-70%-Z has the best thermal conduction efficiency. SEM images in the left column show the morphologies of epoxy and composites. b Top-view infrared images of the LED chips during working, indicating more efficient heat dissipation when GE4-70%-Z is used as TIM. c Digital photographs showing two LED chips integrated with commercial silicone rubber and GE4-70%-Z as TIMs. d Comparison between the temperature increases in the same plot on two chips, depicted by the white dotted circle in b

Full size image

3.3 Fracture Behavior of Epoxy and LSGA/Epoxy Composites

Apart from the high through-plane thermal conductivity, the nacre-like structure also endows the composites with high fracture toughness. The typical force–displacement curves (Fig. 6a) show that epoxy is brittle (curve 1). Because of the enhancement of graphene networks, the force–displacement curve of IGE4 (composite with an isotropic graphene network) is higher than that of epoxy, but is still brittle (curve 2) [63]. The force–displacement curve of GE4 is much higher than those of epoxy and IGE4. With increasing the compression extent, the force–displacement curves show decreases (curves 3–6), but are still higher than those of epoxy and IGE4. Figure 6b shows the initial fracture toughness (KIC) results calculated on the basis of the force–displacement curves. KIC of the IGE4 is ~ 0.64 MPa m1/2, slightly higher than that of epoxy (~ 0.62 MPa m1/2). Fortunately, the lamellar-structured GE4 exhibits an enhanced fracture toughness of ~ 0.88 MPa m1/2, ~ 1.38-fold that of the IGE4 at the same graphene content. Consistent with the force–displacement curves, the KIC decreases gradually to ~ 0.70 MPa m1/2 when the compression extent increases to 70%. As reported previously, when the graphene content is beyond a certain value, the fracture toughness of its epoxy composite would decrease [64]. Fortunately, as compared to neat epoxy, the lamellar structure of LSGA endows its epoxy composite with a high through-plane thermal conductivity of ~ 20 W m−1 K−1 as well as a high fracture toughness at a relatively low graphene content. As shown in Fig. 6c, rising resistance curves (R curve) are calculated to explain the toughness of GE4 and GE4-70%. As the cracks continue to grow, the maximum fracture toughness (KJ) of GE4 gradually increases to ~ 2.06 MPa m1/2 within the American Society for Testing and Materials (ASTM) limit (E1820-13) [65], which is ~ 3.32-fold that of neat epoxy. Although the KJ of GE4-70% is lower than that of GE4, it still reaches ~ 1.06 MPa m1/2, which is ~ 1.71-fold that of epoxy.

Fig. 6
figure 6

a Typical force–displacement curves of epoxy, IGE4, GE4, GE4-30%, GE4-50%, and GE4-70%. b KIC comparison of epoxy and our graphene/epoxy composites. c Rising R curve of maximum fracture toughness versus the crack length. d SEM images show the straight crack propagation. e–g SEM images show the tortuous crack growth of GE4-70%; f and g are enlarged versions of the selected part in e

Full size image

Crack propagation is used to explain the high fracture toughness of GE4-70% (Fig. 6d–g). The crack of epoxy is straight along the tip of the notch (Fig. 6d). In contrast, the crack propagation of GE4-70% is tortuous as shown in Fig. 6e. The enlarged images (Fig. 6f, g) show that the crack deflection, crack branching, and interfacial friction during the crack propagation dissipate a large amount of energy and sequentially endow GE4 and GE4-70% with high fracture toughness [43, 44, 66]. The fracture morphologies of epoxy and GE4-70% illustrate that the graphene lamellae are debonded and even pulled out from the epoxy matrix (Fig. S10), which also contributes to the enhancement in fracture toughness.

4 Conclusions

Lamellar-structured PAAS/GO aerogels are fabricated by bidirectional freezing of PAAS/GO suspensions followed by lyophilization and converted to PI/RGO aerogels by thermal treatment at 300 °C, during which PAAS monomers are polymerized to PI macromolecules, while GO is thermally reduced to RGO. The final graphitization at 2800 °C is crucial for obtaining the lamellar-structured high-quality graphene aerogels, during which PI is carbonized and even graphitized to thermally conductive carbon with the inductive effect of RGO, while RGO is simultaneously upgraded to high-quality graphene by thermally removing its residual oxygen-containing groups and healing its lattice defects. By adjusting the initial mass ratio of PAA and GO, an optimal LSGA with superior thermally conductive capacity is obtained because of its continuous network, densely stacked graphene lamellae, and large graphene sizes. Thanks to the excellent compressibility, the lamellar-structured graphene aerogel infiltrated with epoxy monomer and curing agent could be compressed perpendicular to the lamellar direction to adjust the graphene content in the resultant graphene/epoxy composite. The nacre-like anisotropic composite exhibits different thermal conductivities along three directions, and its through-plane thermal conductivity can be as high as ~ 20.0 W m−1 K−1 at a low graphene content of ~ 2.30 vol%, with a high TCE of ~ 9915% and a record-high specific TCE of ~ 4310%. In addition, the lamellar-structured graphene aerogel also endows epoxy with an enhanced fracture toughness. Our nacre-like graphene/epoxy composite with high through-plane thermal conductivity and fracture toughness demonstrates an insightful avenue for fabrication of high-performance thermal interface materials.