Examining the contribution of factors affecting the electrical behavior of poly(methyl methacrylate)/graphene nanoplatelets composites

In this study, poly(methyl methacrylate) (PMMA)/graphene nanoplatelets (GNPs) conductive composite films with different morphologies were fabricated from the same constituent materials using four fabrication techniques, solution casting (SC), SC followed by hot pressing (SCP), melt mixing followed by SC (MSC), and melt mixing followed by hot pressing (MP). Morphologies of dispersed GNPs and electrical properties in both in-plane and perpendicular direction were investi-gated and compared systematically. The corresponding percolation thresholds ( Φ c ) of the composites varied from 0.42 ± 0.13 vol% to 3.26 ± 0.48 vol%. The conductivities varied up to two orders of magnitude and decreased in the sequence of SC > MSC > SCP > MP. These variations were explained in terms of GNPs size, GNPs orientation, distribution and dispersion state of fillers. The contribution of the above factors in each procedure were discerned individually, the results were discussed and compared with other experimental studies and simulations as well.


| INTRODUCTION
The incorporation of conductive fillers into nonconductive polymer matrices allows to obtain a new class of electrically conductive polymer composites (CPCs). Traditional carbon-based fillers such as carbon nanotubes 1 and carbon black 2 make excellent candidates for conductive composites. While low percolation thresholds have been observed, achieving a high ultimate conductivity requires a greater loading of filler, and this often makes the composite brittle. 3 Therefore, graphene has attracted significant scientific attention as a potential alternative conductive filler. Graphene is a two-dimensional sheet of carbon atoms, sp 2 -bonded into a hexagonal arrangement. 4 It has an exciting combination of properties, especially its superb electrical conductivity (up to 10 8 S m −1 ). 5,6 There is now a family of graphene-related materials, such as pristine graphene, 7-12 graphene oxide, [13][14][15][16][17][18] reduced graphene oxide, [19][20][21] and graphene nanoplatelets (GNPs). [22][23][24][25] Typically, GNPs are much cheaper than other forms of graphene and are commercially available on the tonne scale. 26 The effects of filler orientation on the conductivity of the composite have been extensively studied recently, and anisotropic electrical properties have been reported. The orientation of fillers could happen by using conventional manufacturing methods, [27][28][29][30] novel manufacturing methods, [31][32][33] or applying an external electric or magnetic field. [34][35][36][37] Wu et al. 28 found the compression molded samples showed slightly orientation of graphene and higher electrical conductivity with respect to the injection molded ones. For compressed samples, the inplane percolation threshold (Φ c ) was around 1 wt% and the perpendicular Φ c was 3 wt%. For injected samples, the inplane Φ c was 10 wt% while that of perpendicular direction was higher. Wang et al. 31 prepared graphene aerogels using a novel unidirectional freeze casting method, the obtained highly aligned, porous structure of aerogels caused over an order of magnitude difference in conductivity of epoxy composite. Wu et al. 34 used electric field to align graphene in an epoxy polymer, the conductivity of the CPCs in the alignment direction of the graphene was consistently 2-3 orders of magnitude higher than both that along the transverse direction and that of the CPCs containing randomly oriented graphene. The corresponding Φ c were 0.22 and 0.52 vol% for CPCs with aligned graphene and randomly oriented graphene, respectively. However, few of these studies reported the effect of other factors in addition to filler orientation, such as shape and size of fillers, polymer-filler interactions, dispersion and distribution state of fillers.
If the fillers have a large contact surface and aspect ratio, the nanostructures make it possible to form conducting networks much more efficiently than fillers with a spherical shape. 38,39 Stronger interfacial interactions or compatibility between polymer and fillers are expected to result in more efficient stress transfer to the fillers during compounding, thus breaking up more effectively the filler aggregates. 40 Compatibilization strategies are usually used to establish stronger interactions between polymer matrices and fillers. [41][42][43][44][45] These strategies ultimately help achieve effective dispersion in the polymer matrices. Dispersion and distribution of fillers not only depend on polymer-filler interactions, but also strongly depend on the processing, including compounding techniques and equipment, time of mixing, and applied shear. 42 In situ polymerization, 46,47 melt mixing 48,49 and solution mixing 50-52 the three most commonly used blending techniques for the preparation of CPCs. From those comparisons of different mixing methods in terms of particle dispersion and properties of produced composites, 27,53,54 it has been found solution mixing facilitates material transport in low viscosity solvents and facile polymer diffusing between the exfoliated graphene sheets, resulting in further exfoliation of the graphene sheets, thus lower Φ c and greater conductivity of the CPCs were generally observed relative to melt blending method. Novel mixing method, such as microsphere rolling transfer, was reported to significantly increase the electrical conductivity, much higher than traditional solid mixing methods and approaching to the best from solution mixing of graphene fillers. 55 However, these researches only dealt with the effect of mixing method on either in-plane or perpendicular electrical properties rather than both.
It could be concluded that, on one hand, the influence of the above factors on the electrical properties of CPCs are complicated and may vary from CPC system to system. On the other hand, these factors could play a role simultaneously in processing. Although the electrical conductivity of composites could be predicted through analytical and numerical studies, it is difficult to discern the contribution of each factor using experimental techniques. 56 The present work aims to explore the impact of the above factors individually (orientation of fillers, filler size, dispersion and distribution state of fillers, etc.) on both inplane and perpendicular electrical properties of graphenebased polymer composites. Solution mixing and melt compounding were used to disperse GNPs into poly(methyl methacrylate) (PMMA) matrix, and solution casting (SC) and compression molding were selected to fabricate the CPC samples. With simple experimental design, processing, morphologies, and properties were presented.
All materials were dried at 80 C in a vacuum oven for at least 12 h prior to processing. As shown in Figure 1, the PMMA/GNPs composites were prepared via solution mixing and melt mixing, the films were fabricated via casting and hot pressing. The solution mixing was carried out under ambient conditions (around 23 C, 55% RH). GNPs of various loading were sonicated in THF for 30 min to yield a homogeneous dispersion, and then PMMA pellets were added into the GNPs/THF suspensions. The concentration of polymers was kept at about 4 wt% in this work. The final concentration of GNPs in PMMA/GNPs composites was from 0.3 to 4.0 vol%. Next, the PMMA/GNPs/THF mixtures above were stirred 24 h. Finally, these mixtures were cast onto glass plates and dried under ambient conditions for 48 h (SC sample). Further hot press on the SC sample was employed under vacuum at a temperature of 200 C and a pressure of 200 bar for 5 min (SCP sample). For melt mixing, GNPs and PMMA were mixed in an internal kneader PolyDrive (Haake, type 557-8310) at a temperature of 200 C and a rotation speed of 60 min −1 for 10 min. The designed GNPs fractions were from 0.5 to 5.0 vol%. After granulation, the pellets were dissolved into THF followed by casting (MSC sample) or were hot pressed into films (MP sample). The thickness of all the samples was controlled to be about 0.2 mm.

| Characterization
The actual GNPs weight fractions of the melt mixed composites were determined by thermogravimetric analysis (TGA) under a nitrogen atmosphere, with a heating rate of 10 C/min. The relationship between GNPs volume fractions and weight fraction can be described by Equation (1), vol = wt=ρ GNPs wt=ρ GNPs where ρ GNPs = 2.2 g cm −3 and ρ PMMA = 1.19 g cm −3 . The result of TGA is shown in Figure S1, Sample code, volume fraction as well as weight fraction of GNPs are summarized in Table 1. Investigation of the macro-scale distribution of GNPs in PMMA matrix was observed with an optical microscopy (OM) (Leitz, Orthoplan P). To determine the geometrical size of GNPs in the composites, samples were redissolved in THF to form dilute solutions, followed by casting on glass plates and observing with the light microscope. The images obtained were then analyzed using the JMicrovision image analysis freeware. At least 500 GNPs per concentration were measured.
Scanning electron microscopy (SEM) (Carl Zeiss Microscopy, Germany) was conducted to investigate the morphologies of the composite films, operating at an accelerating voltage of 3 KV. The composite films were previously fractured at room temperature and the crosssections were coated with a gold layer using Sputter Coater S150B from Edwards.
The electrical resistances R of composites both in in-plane direction and through-plane direction were measured under ambient temperature and humidity (around 23 C, 55% RH), using a Keithley 6487 Pico ammeter at a constant voltage (1 V) based on a homemade setup as shown in Figure 2. The electrical conductivity σ was then calculated as: where R is the electrical resistance of the sample, d is the length of the sample parallel to the direction of the voltage or the distance between two electrodes, S is the area of the sample perpendicular to the direction of the voltage.

| GNPs size
The OM image and lateral area distribution of as received GNPs and GNPs in composites prepared from different routes are shown in Figure 3. The uncertainty Δ represents the 95% confidence interval for the GNPs size.
where N is the number of units studied, N ≥500 in this work, taking from five images for five samples. SD is the standard deviation. As shown in Figure 3(a), the as-received GNPs had an average lateral size of 1137 μm 2 with the most size between 200 and 400 μm 2 . Regarding GNPs as circles, an average diameter of 38 μm could be obtained. The size of GNPs in SC 2.0 sample ( Figure 3(b')) was comparable while that in MSC 1.9, SCP 2.0, and MP 1.9 sample (Figure 3(d'), (e')) showed obvious smaller size. This indicates that SC method preserved the GNPs better while the GNPs could be efficiently broken up during melt compounding with the PMMA matrix or during the hot pressing. In other words, simple mechanical shearing at molten state can facilitate the exfoliation of the GNPs into smaller dimension. 28,57 The size reduction presumably caused by the interaction of GNPs-PMMA, interaction of GNPs-GNPs and processing equipment surfaces.
In order to investigate the relation between GNPs size and GNPs content, GNPs size of composites with various filler loading was also measured and the mean values are shown in Figure 4. For SC composites, SC 0.5 had a size of 1178 ± 124 μm 2 , and all showed negligible difference at different concentration. The slight increasement at 3.0 vol% (1281 ± 152 μm 2 ) might be ascribed to the formation of some agglomerations of GNPs as the concentration increased. 58 Interestingly, in the case of MSC sample the rise became more remarkable. As GNPs concentration increased from 0.9 to 3.9 vol%, the size of GNPs increased from 522 ± 60 μm 2 to 868 ± 180 μm 2 . Presumably some agglomerations of GNPs also formed at higher GNPs concentration during melt compounding, which could not be completely broken when dissolved. SCP sample had a similar size of GNPs (1110 ± 163 μm 2 ) to SC sample at a content of 0.5 vol%, which dropped to 681 ± 71 μm 2 as it increased to 1.0 vol% and almost kept constant with further loading increasing. Unlike MSC sample, with GNPs increasing, the size of GNPs of MP composites decreased from 613 ± 83 μm 2 (0.9 vol%) to 550 ± 81 μm 2 (3.9 vol%). The reduction of GNPs size in SCP and MP sample could be attributed to the interaction of GNPs-PMMA and GNPs-GNPs improving due to more steric hindrance imposed by more particles during the compression process. 59 3.2 | Distribution, dispersion and orientation of GNPs in the PMMA matrix The surface of the composite films prepared with four different processing methods was observed with optical microscopy. Figure 5 shows macroscopic distribution levels of GNPs in the composite samples in the in-plane direction. Strong surface segregation of GNPs in transparent PMMA matrix was evident in the case of the SC and MSC sample while that in SCP and MP sample was not observed. Thus, this phase separated morphology appears to have been induced during the SC process, independent on the compounding method. During solvent evaporation, the intrinsic incompatibility between GNPs and PMMA may drive the segregating process. 60 On the contrary, SCP and MP samples showed well-distributed GNPs due to the shear forces during compression. 61 The morphologies of the cross-section of composite films were observed with SEM and the result is shown in Figure 6. The distribution at micro-scale of GNPs through the thickness is clearly demonstrated. Some isolated graphene clusters could be discerned as shown by the rough parts in the images. In Figure 6(a), MSC 1.9 showed layered structure with upper PMMA-rich layer and under GNPs-rich layer. The layered structure indicates a poor compatibility between GNPs and matrix in a solution state. The loss of compatibility relative to SC sample might be attributed to the high temperature or shearing during melt mixing. With poor compatibility with matrix, most GNPs were deposited toward the bottom due to the gravity effect. 62 As the GNPs content increased to 3.9 vol% (Figure 6(b)), the polymer layer became thinner. For SC, MP samples ( Figure S2) and SCP samples (Figure 6(c), (d)), this layered structure did not occur, indicating better compatibility between the matrix and fillers.
It can also be seen from Figure 6 that most of the GNPs tended to orient along the in-plane direction for all samples. Besides, the level of the GNPs orientation varied with the GNPs concentration. As shown in Figure 6(d), the alignment of GNPs along in-plane direction in SCP 4.0 was more severe than that in SCP 3.0 ( Figure 6(c)). This self-alignment phenomenon could be explained by the steric hindrance and π-π interactions between the GNPs as well as the excluded volume concept. 27,28 High magnification SEM images of composite films are displayed in Figure 7, further demonstrating the GNPs dispersion at nano-scale (or exfoliation) for the four different processing methods in detail. SC 2.0 sample (Figure 7(a)) had large zones covered by loose GNPs flakes and had homogeneous dispersion. In Figure 7(b), several GNPs flakes stacking together is visible for MSC 1.9 due to the abovementioned layered structure. Similarly, SCP 2.0 (Figure 7(c)) also shows tighter stacks of GNPs, while MP 1.9 (Figure 7(d)) shows tightest and least homogeneous GNPs clusters. This is due to strong π-π bonding between graphene layers without solvent-aided dispersion, as a result, large parts of PMMA matrix can be seen from the image. 63,64 From the findings, it is considered that SC process had the best exfoliation of GNPs. It should be noted that the distribution, dispersion and exfoliation are different terms to describe graphene states. "Distribution" describes the way the individual sheets or their agglomerates fill the matrix at micro-and macro-scopic scales. "Dispersion" indicates GNPs are agglomerated or not, dispersion at nano-scale reflects the "exfoliation" or "stacking" state of graphene sheets in the matrix. 65 For CPCs, the ideal state of fillers in the matrix is bad distribution and good dispersion. The conductive network can easily form from bad distribution due to the high concentration of conductive particles in a local position such as the segregation in SC and MSC sample, while the good dispersion means more conductive particles can be released from the filler aggregates to construct conductive network. 61 The summary of the results of morphology analysis was listed in Table 2.

| Electrical conductivity and percolation behavior
The electrical conductivity of CPCs is strongly dependent on the volume fraction of conductive fillers. Percolation threshold (Φ c ) is a certain concentration at which the electrical conductivity of CPCs increases by several orders of magnitude and the insulator/conductor transition is achieved in the percolation process. Based on the conductivity data, McLachlan's general effective medium (GEM) equation 66 was found to be able to describe successfully the conductivity data in a broad range of filler concentrations, that is, below and above Φ c as well as the transition region, for many conductive composites. This equation is as follows: where σ m , σ c , σ f are the conductivities of the PMMA matrix, the composite and the GNPs, respectively. Φ is the volume fraction of fillers and Φ c is the percolation Note: + denotes level of the item, more +, higher level. " means increase, # decrease and the same as the left item. Abbreviations: GNP, graphene nanoplatelets; MP, melt mixing followed by hot pressing; MSC, melt mixing followed by SC; SC, solution casting, SCP, SC followed by hot pressing.
threshold. The values of the exponent s and t are generally taken to be 0.87 and 2, respectively.
The GEM Equation (4) was used to obtain the percolation threshold Φ c . σ m (2.28 × 10 −13 S cm −1 ) was measured on a pure PMMA film cast or compressed under the same conditions as the other samples. σ f is 800-1100 S cm −1 stated by the manufacturer. The experimental conductivity data in both in-plane (denoted as -I) and perpendicular direction (denoted as -P) as a function of GNPs volume fraction were described. Total of 9 measurements were conducted to obtain the mean conductivity and standard deviation as shown by vertical error bars. As shown in Figure 8, the theoretically calculated conductivities (dashed lines) were close to the experimentally determined conductivities (points).
Based on the geometry of conductive fillers, Li and Jim 67 developed an analytical model based on average interparticle distance (IPD) approach to predict the percolation threshold of polymer nanocomposites containing 3D randomly distributed disc-shaped nano particles.
The equation is given as: where t is the filler thickness, D is the filler diameter, and D IP denotes the IPD. The critical D IP values are varied between 10 nm and 1 μm, the effect of D IP criterion on percolation threshold becomes negligible as far as the aspect ratio is high enough (i.e., D/t ≥about 500).
It has been confirmed that the theoretical Φ c value calculated from Equation (5) was in accordance to the determined Φ c by considering a proper average lateral dimension and thickness of GNPs. 68 In this work, Φ c generated from GEM Equation (4) is regarded as experimental data and is listed in Table 3. Equation (5) yields the dependence of Φ c on the diameter D, thickness t, and IPD by considering the GNPs 3D randomly distributed. D was between 25 and 40 μm for composites prepared with different methods (calculated from Section 3.1). If GNPs distributed without any further exfoliation or stacking as received, then t could be considered 3.35 nm (interlayer space: 0.335 nm, 10 layers) as stated by the manufacturer. Here, the D/t ≥500, D IP can be neglected mathematically. As an isotropic GNPs orientated system, the model predicts a Φ c between 0.18 and 0.28 vol%.
It is noted that the experimental Φ c were all higher than that predicted from Equation (4). This reveals the discrepancy between the experimental system and the ideal model based on randomly distributed GNPs. A major reason is, the actual GNPs thickness would be larger than 3.35 nm used for predicting Φ c , as the stacked GNPs flakes shown in the SEM images, and the thickening of GNPs increased Φ c . Although the actual thickness of each case cannot be speculated simply with Equation (5) since there were other differences in the morphology of composites between the model and experiment, the relationship between Φ c and GNPs diameter and GNPs thickness could further help investigate the contribution of GNPs size and GNPs stacking in the following sections.

| Contribution of GNPs size and layered structure to the electrical properties
Considering GNPs size and layered structure in MSC sample as the two main factors determining the electrical conductivities and percolation behaviors of SC and MSC sample, differences between them could be explained. As demonstrated by Equation (5), the composite with a larger diameter has lower percolation threshold. In the simulated result of PMMA/GNPs composite by Zabihi and Araghi, 56 lower percolation threshold as well as lower conductivity were also obtained for larger GNPs. In agreement with the simulations, SC sample with larger GNPs had a lower Φ c,P (0.42 ± 0.13 vol%) than MSC sample (0.77 ± 0.18 vol%). Besides, a lower log σ I was also found for SC sample. However, the conductive behavior of MSC sample was also influenced by the layered structure.
Only if GNPs formed a percolated network in the top PMMArich layer, the perpendicular direction started to conduct. 62 That is another reason why MSC sample had a greater Φ c,P than SC sample. Similarly, the log σ P of MSC sample was lower than that of SC sample at the beginning but approached that of SC sample as the concentration of GNPs increased to 3.0 vol% since the matrix-rich layer got thinner. It seems the effect of layered structure dominated this process. On the other hand, more conductive paths in the inplane direction would be formed in the bottom area compared to uniformly dispersed systems. This is another factor leading to a higher log σ I of MSC sample in addition to their smaller filler size. Similarly, a smaller amount of GNPs would be required to achieve percolation threshold since it became conductive as long as the bottom GNPs-rich layer percolated. As a result of the competition between positive effect of layered structure and negative effect of smaller GNPs size, MSC sample eventually had a slightly lower Φ c,I (0.67 ± 0.11 vol%) than SC sample (0.8 ± 0.25 vol%). This finding also indicates that the difference in GNPs size had less significant influence on the electrical properties of composites than layering effect.

| Contribution of the segregation of GNPs to the electrical properties
Comparing the percolation thresholds of SC and SCP composites, Φ c,I and Φ c,P of SC sample were 0.8 ± 0.25 vol% and 0.42 ± 0.13 vol%, respectively, while those of SCP sample were 2 ± 0.03 vol% and 0.89 ± 0.16 vol%, respectively. In addition to the difference in GNPs size which has less profound impact on the electrical properties, this could be mainly explained by the segregation of GNPs in the SC sample, which disappeared after hot pressing for SCP sample. In other words, the phase segregation effect reduced Φ c,I and Φ c,P of about 1.2 and 0.47 vol%, respectively. This indicates that GNPs segregation was a benefit for conductive network, facilitating the percolating process at lower amount of GNPs. It could also be concluded that it had a more significant promotion effect on the formation of conductive pathways in inplane direction. With respect to the electrical conductivity of SC and SCP sample (when Φ>Φ c ), both log σ I and log σ P of SC sample were about one order of magnitude higher than those of SCP sample, and the difference between them remained almost the same as the GNPs content increased. Here the segregation of GNPs means a poor distribution resulting from the incompatibility between matrix and fillers. The better electrical properties due to this poor distribution was in agreement with some literatures 40,41 where a slight increase in Φ c was observed as the compatibilization method applied. This is ascribed to an increased number of individually dispersed platelets and a loss of interconnectivity.

| Contribution of the stacking of GNPs to the electrical properties
In comparison to SCP composite (Φ c,I = 2 ± 0.03 vol%, Φ c,P = 0.89 ± 0.16 vol%.), MP sample showed higher percolation threshold values (Φ c,I = 3.26 ± 0.48 vol%, Φ c, P = 2.32 ± 0.53 vol%). The filler size of SCP and MP composites is similar and no segregation of GNPs were found for both, therefore the difference in Φ c was mainly resulted from the different GNPs stacking states. This is in consistent with the morphology analysis which showed obviously tight and less homogeneous GNPs clusters for MP sample. According to Equation (5), Φ c is directly proportional to the thickness t in a 3-D isotropic conductive system. Considering the stacking of GNPs as the increase of GNPs thickness, the rise of Φ c for MP sample with respect to SCP sample could be explained from the view of the simulation. The level of increase in Φ c,P was higher than that in Φ c,I , this implies that stacking of GNPs caused more destruction of the conductive pathways in perpendicular direction. MP sample showed lower log σ I and log σ P than SCP sample in Figure 8. The differences between the two composites were greater before the insulator-to-conductor transition of MP sample. However, as the volume fraction of GNPs increased above both percolation thresholds, the lower log σ I and log σ P of MP sample became closer and closer to those of SCP sample. From these findings, it can be concluded that the GNPs stacking had less impact on electrical conductivity than on percolation threshold.

| Contribution of GNPs orientation to the anisotropic electrical properties
It can be seen from Figure 8 that the in-plane conductivity was about one or two orders of magnitude higher than the perpendicular direction for the same composites, which is consistent with the result of preferential inplane orientation of GNPs stacks showing in SEM images. The percolation regions of in-plane direction where a steep increase in conductivity was observed were almost at higher filler loadings.
In order to quantitatively investigate the anisotropic electrical behaviors, the resulted Φ c from Equation (4) is shown in Figure 9(a). Opposite to the result in other researches, 28,34 Φ c,I >Φ c,P was found in this work for SC, SCP and MP sample as shown in Figure 9(a). It might be ascribed to the significant difference in the geometry of GNPs and the conductivity measuring distance. 69 It was found for a PMMA composite 62 filled with carbon fiber having a large aspect ratio that the Φ c,I and Φ c,P were 1.80 ± 0.40 vol% and 0.27 ± 0.06 vol%, respectively. Although the in-plane orientation for GNPs was favorable during processing, it was much more difficult to form a continuous conductive network in a 25 mmlong in-plane measuring distance than in a 0.2 mm perpendicular measuring distance, with a mean diameter of GNPs between 25 and 40 μm. Unlike other samples, MSC sample had a higher Φ c,P than Φ c,I . This could be attributed to the layered structure. The upper matrix-rich layer hindered the formation of connectivity network through the thickness, while the under GNPs-rich layer facilitated the connection along the inplane direction.
In addition, anisotropy of the electrical conductivity was defined to be the ratio of experimental in-plane conductivity to perpendicular conductivity.
The result is shown in Figure 9(b). The anisotropy of SC, SCP and MP sample increased with the GNPs fraction increasing as also found in other studies, 29,32,70 relating to the faster increase rate of σ I than σ P . This was caused by the increase in orientation level with GNPs fraction. However, the anisotropy of MSC sample decreased, since the layered structure became less with the filler content increasing. The result is in agreement with the morphology investigation in Figure 6. In this work, four processing methods named SC, SCP, MSC, MP were used to fabricate PMMA/GNPs conductive composite films. The morphologies, both in-plane and perpendicular direction conductivities σ and percolation threshold Φ c of all the composites were investigated. The corresponding percolation thresholds of the composites varied from 0.42 ± 0.13 vol% to 3.26 ± 0.48 vol% and the conductivities varied up to two orders of magnitude. Among the four routes, the ability of improving electrical properties decreases in the sequence of SC>MSC>SCP>MP. The contribution of GNPs size, distribution state (GNPs segregation and layered structure), dispersion state of fillers (stacking) and GNPs orientation to the electrical behavior of composites were revealed individually. Furthermore, the results were discussed and compared with other experimental studies and simulations. This work provide a deeper insight into understanding the individual effect of each factor and a systematic guide for the preparation of graphene-based polymer composite with superior electrical performances. Continuous research would further avoid the stacking by hybridizing GNPs with other fillers.