Enhancement of barrier properties of cement mortar with graphene nanoplatelet

Cement and Concrete Research 76 (2015) 10–19

Contents lists available at ScienceDirect

Cement and Concrete Research journal homepage: http://ees.elsevier.com/CEMCON/default.asp

Enhancement of barrier properties of cement mortar with graphene nanoplatelet Hongjian Du, Sze Dai Pang ⁎ Department of Civil and Environmental Engineering, National University of Singapore, 117576, Singapore

a r t i c l e

i n f o

Article history: Received 7 August 2013 Accepted 6 May 2015 Available online xxxx Keywords: Diffusion (C) Dispersion (A) Durability (C) Permeability (C) Pore size distribution (B)

a b s t r a c t The transport properties of cement mortar with graphene nanoplatelet (GNP) are investigated experimentally in this study. GNP, a low cost carbon-based nano-sheet, was added to mortar at contents of 0, 2.5, 5.0 and 7.5%, by weight of cement. The water penetration depth, chloride diffusion coefficient and chloride migration were determined for cement mortar with GNP and compared with plain cement mortar specimens. Test results showed that the addition of 2.5% GNP can cause significant decrease of 64%, 70% and 31% for water penetration depth, chloride diffusion coefficient and chloride migration coefficients respectively. The reduced water and ions ingress can be partially attributed to a reduction in the critical pore diameter of about 30%. This refinement of the microstructure by the GNP is validated by the mercury intrusion porosimetry (MIP) results. The impermeable GNP also contributes to the reduced permeability due to the increased tortuosity against water and aggressive ions ingress. © 2015 Elsevier Ltd. All rights reserved.

1. Introduction Durability remains one of the most important properties of concrete, especially for structures exposed to aggressive environments or severe conditions [1–3]. The improvement in the durability performance of concrete has been widely accepted by the community as a means to reduce the life-cycle cost of an infrastructure especially in the maintenance during its service life [4]. The durability of concrete structures is strongly influenced by its transport properties to the harmful agents such as water, CO2, chloride, etc [5]. Existing durability enhancement methods that are commonly used include lowering water–cement ratio with the aid of superplasticizers, adding supplementary cementitious materials and using chemical admixtures [6–8]. In recent years, there has been rising interest in the use of nanoparticles in building materials to enhance mechanical performances and to create multifunctional capabilities [9–13]. These nano-particles can fill the voids in the cement paste, leading to lower porosity and higher strength [10–12]. The research on using nano-particles to improve the transport properties of cementitious composites is, however, limited. Ji [14] reported that the nano-SiO2 can fill the voids in cement paste and thus reduce the water permeability of concrete. Gaitero et al. [15] used nano-SiO2 to improve the resistance to calcium leaching of cement paste which can be attributed to the reduced porosity in the cement paste, increased length of the silicate chains and pozzolanic reaction of nano-SiO2. At the same time, only few studies were carried out ⁎ Corresponding author. Tel.: +65 65162799; fax: +65 67791635. E-mail address: [email protected] (S.D. Pang).

http://dx.doi.org/10.1016/j.cemconres.2015.05.007 0008-8846/© 2015 Elsevier Ltd. All rights reserved.

to study the transport properties of mortar and concrete, which are cement composites containing aggregates. Zhang and Li [16] reported that the addition of nano-SiO2 and nano-TiO2 can refine the pore structures and therefore reduce the permeability of concrete to chloride. Kong et al. [17] recently studied the influence of agglomeration of nano-SiO2 on the performance of cement mortar and they reported that despite the existence of the weak zone between the cement paste and the agglomerates, and the higher porosity of agglomerates, the nano-SiO2 can still be effective in blocking the ingress of chloride ions. So far, nano-particles that have been introduced into cementitious materials include nano-SiO2, nano-TiO2, nano-CaCO3, and nano-C such as carbon-nanotube and carbon-nanofiber [18]. Another form of carbon-based nano-material is the graphene nanoplatelet (GNP) which has been widely studied in polymer nanocomposites. GNP consists of several layers of graphene with a total thickness of less than 100 nm and a diameter of several micrometers (as shown in Fig. 1(a)). Jang and Zhamu [19] reported that GNP reinforced nanocomposites combine the benefits of good mechanical properties and impermeability. According to Compton et al. [20], the layered structure of GNP is effective in inhibiting the transport processes for gas and fluid in the host matrix due to increased tortuous paths. Another appealing advantage of GNP is the low cost when compared with other carbon nanoparticles, such as CNT or mono-layer graphene, whose large scale production is significantly more expensive [21,22]. Despite the numerous advantages of GNP and its wider adoption in polymer based composites, its potential has yet to be tapped upon in cement composites until recently [23–35]. The introduction of GNP into cement composites was found to increase the electrical conductivity by 2–3 orders of magnitude

H. Du, S.D. Pang / Cement and Concrete Research 76 (2015) 10–19

11

use of two-dimensional graphene as a highly effective barrier to enhance the barrier properties of cement composites. 2. Materials and methods 2.1. Materials Ordinary Portland cement (OPC) was used in this study, with chemical composition shown in Table 1. The Blaine fineness of OPC is 393 m2/ kg. Natural sand with density of 2.65 (saturated surface dry) and fineness modulus of 2.95 was used. GNP was exfoliated from surfaceenhanced expanded graphite flake (Grade A3775, Asbury Graphite Mills, Inc.). Its physical properties are shown in Table 2. The mix proportion for water: cement: sand was selected as 0.485:1:2.75, by weight according to ASTM C 109 [26]. GNP was added at contents of 0%, 2.5%, 5.0% and 7.5% by weight of the cement corresponding to 0%, 0.6%, 1.2% and 1.8% by volume of mortar, respectively. The GNP particles will form precipitates in water which will remain as agglomerates when the suspension is added to mortar mixture and this will negate the benefits of nano-particle addition. A naphthalene sulfonate based superplasticizer (SP) (Darex Super 20, WR Grace) is used to disperse the agglomerates and stabilize the GNP particles in aqueous solution. The added SP could be absorbed at the GNP surface, surrounding them with negative charges and repel each other. This uniformly dispersed aqueous suspension of GNP with SP was mixed with cement and sand to create mortar with GNP. Different amounts of SP have been tested for the flowability of the cement mortar with GNP and the amount of SP is kept at a constant 50% of the weight of GNP in this study to ensure good consistency. In comparison with the recommended dosage of 400–2500 ml of SP to disperse 100 kg of OPC, SP is used at a higher dosage in this study to disperse the GNP which has a surface area 60 times (Table 2) larger than that of OPC. Even at 2.5% GNP, the total surface area of GNP is about 1.5 times of that of OPC. Thus, the use of 1500 ml SP for every 2.5% GNP lies within the expected range of 600–3750 ml SP based on surface area requirements for dispersion of GNP. With the above mentioned SP, all the mortar mixtures had similar flowability (with mini-slump diameter of 150 ± 10 mm). 2.2. Casting procedure

Fig. 1. SEM images of GNP A3775 (a) after sonication, (b) at the fractured mortar surface, and (c) clustering at 7.5% mortar.

and diminishes its sensitivity on the moisture content of the specimen [23]. This enhanced conductivity of mortar infused with GNP can also be exploited for creating self-(damage) sensing cement composites and simple closed-form solutions have been derived for predicting the damage in this conductive cement nanocomposite [24,25]. The first study on the transport properties of GNP nanocomposites in cement matrix medium is reported in this paper. Chloride diffusion and migration tests and water permeability tests were carried out according to the various test standards to investigate the main transport mechanisms of diffusion, migration and permeability respectively. The results for cement mortar with GNP are compared with plain mortar and the enhancement in the barrier properties brought about by the introduction of GNP is explained by the changes in pore structure, which are revealed via the mercury intrusion porosimetry (MIP) test, and the increase in tortuosity against water and chloride ion ingress, which are explained with analytical models. This work can shed light on the

Prior to casting of mortar mixture, GNP was first ultra-sonicated with the aid of water and the SP. The desired amount of GNP was added to the mixing water together with dispersant and then handstirred for 1 minute. A high power ultra-sonication horn was used to disperse this suspension for 1 hour in a water bath to cool down the horn. According to previous literature [27], sonication technique is typically considered as a nondestructive process for the exfoliation of graphene which preserves the basal plane of GNP flakes. However, if the sonication time is excessively long (for example 5 to 10 hours as determined by Xia et al [28]), graphene can suffer from size reduction and damage. The efficiency of dispersing GNP without sonication in water, in water with SP, and the effect of sonication in water with SP can be observed via visual inspection (See Fig. 2.). In water, sedimentation of GNP at the bottom of the bottle was noticed after 2 hours. In contrast, GNP Table 1 Chemical composition of OPC. Chemical composition

%

SiO2 Al2O3 Fe2O3 CaO MgO SO3 Na2O K2O

20.8 4.6 2.8 65.4 1.3 2.2 0.31 0.44

H. Du, S.D. Pang / Cement and Concrete Research 76 (2015) 10–19

with soniction

Properties

Values

Density (g/cm3) Surface area (m2/g) Diameter (μm) Thickness (nm) Aspect ratio Purity (%)

2.26 24 8.0 37 210 98.0

could be dispersed better in the aqueous solution with SP, particularly with the help of sonication after 2 hours. After 24 hours, all the GNP particles precipitated even in the presence of SP. Although sedimentation could also be noted for the sonicated GNP, this dispersion technique has demonstrated its advantages over others and was employed in this study. Particle size distribution for GNPs before and after sonication were measured respectively and shown in Fig. 3. The distribution has been shifted to a finer size, indicating that the agglomerates have been reduced. The disintegration and dispersion of the GNPs by the sonication process allow the SP stabilizing the aqueous suspension. After sonication, the GNP suspension was added to the mixture of cement and sand in a Hobart mixer. The mixing lasted for 3 minutes. The mortar was poured into steel molds and compacted on a vibration

Intensity, %

Table 2 Physical properties of GNP A3775.

without sonication

40

100

30

75

20

50

10

25

0

Cumulative intensity, %

12

0 1000

10000

Size, d.nm Fig. 3. Particle size distribution of GNP without and with ultra-sonication (measured by dynamic light scattering on Malvern Zetasizer Nano ZS).

table. After casting, all the specimens were covered with a plastic sheet to prevent water loss and demolded after 24 hours. The mortar specimens were cured in a fog room until the testing age.

2.3. Test methods and specimens

Fig. 2. Optical photographs of GNP dispersion after (a) 2 hours and (b) 24 hours in (from left to right) water, SP + water and SP + water with sonication.

For each mortar mix, three 50 × 50 × 50 mm cubes, eight 40 × 40 × 160 mm prisms, three Φ100 × 50 mm discs and three Φ100 × 200 mm cylinders were prepared. At 28 days, compressive strength was determined according to ASTM C 109 [26]. Four of the prisms were tested for flexural strength using three-point bend test according to ASTM C 348 [29]. MIP was used to characterize the distribution of pore sizes in each mortar mix from the fracture surface of mortar at the age of 28 days. Micromeritics Autopore WIN9400 MIP machine was used in this study with a maximum pressure of 412.5 MPa. The samples were kept in an oven at 40 °C for 2 days before MIP testing. Two Φ100 × 200 mm cylinders were tested for their water permeability according to BS EN 12390-8 [30]. The cylinders were coated with epoxy on the bottom and side surfaces to waterproof them and were then applied a water pressure of 0.75 MPa on the top surface for 14 days. After that, the cylinders were split to measure the average depth of the water penetration front. The remaining four prisms and the cylinder were tested for their chloride diffusion according to NT BUILD 443 [31]. The cylinder was wet cut from the middle portion after curing in water for 27 days and epoxy was applied on the top and side surfaces. These two cylinder halves and the four prisms were immersed in a salt solution (185 g NaCl per liter). At the immersion age of 3, 7, 14 and 28 days, one prism was taken out, broken in flexure and sprayed with 0.1 N silver nitrate solution on the fractured surface to determine the chloride ion penetration depth. At the immersion age of 28 days, the cylinder halves were taken out to investigate the chloride content profile by sampling at depths of 5, 15, 25, 35 and 45 mm. The chloride content was determined at each depth by combining 2 powdered samples that were extracted using a 20 mm diameter rotary drill from each cylinder half. After being oven dried at 105 °C, the samples were ground to pass through a 75 μm sieve. Total chloride (acid soluble) content by weight of material was determined as per BS 1881-124 [32]. The three Φ100 × 50 mm discs were tested for their chloride migration at 28 days according to NT BUILD 492 [33]. A constant voltage of 20 V was applied to all discs for 6 hours. The initial and final temperatures were also recorded. After that, the specimens were split axially and sprayed with 0.1 N silver nitrate solution to measure the chloride penetration depth.

H. Du, S.D. Pang / Cement and Concrete Research 76 (2015) 10–19

3. Results and discussion 3.1. Pore size distribution

0.12 0.09

(a)

0GNP+0SP (Ref.) 2.5GNP+1.25SP 5.0GNP+2.50SP 7.5GNP+3.75SP

0.06 0.03 0.00 0.12

(c)

0.09 0.06 0.03 0.00 1E-3

0.01

0.1

1

Pore diameter (Dp), m

large influence on the transport properties of cement-based materials. With the addition of as little as 2.5% GNP with 1.25% SP, the critical pore size is reduced by more than 30% and this is solely due to the effect of GNP since the addition of the same amount of SP only does not have any effect on critical pore size. Considering the nanoscale thickness of GNP particles, they can provide nucleation sites for cement hydration products. The hydration products grow toward the water filled voids and become more uniformly distributed spatially, leading to a more refined microstructure. The experimental results of GNP on cement hydration are discussed in another paper. Similar microstructural refinement has also been observed for other nano-particles, like CNT [34–37], nano-TiO2 [11] and nano-SiO2 [12]. While further addition of GNP content beyond 2.5% does not further reduce the critical pore size, it has an effect of reducing the much larger macropores resulting in an average macropore size of 81.4 nm for GNP content of 7.5% as compared to an average macropore size of 117 nm for GNP content of 2.5% as shown in Table 4. This is in contrast with cement mortar with the same amount of SP only which resulted in a highly porous mix with large amount of air voids (N10,000 nm). The added GNP would reduce the free SP amount in pore water and block the pathways for bleeding water caused by addition of SP in the same cement mix. The introduction of GNP into the cement matrix has little effect on both the large and small mesopores which is reflected by less than 5% change in the average pore sizes of these mesopores. The refinement of the pore structure especially on the macropores is also reflected in the decrease in the average and median pore diameters and these effects, arising from the introduction of GNP, are expected to enhance the barrier properties of cement mortar which are investigated in the following sections. It should be noted that MIP technique has an intrinsic drawback in deducing the true volume of pores with each size due to the existence of ink-bottle effect of the pore system [38]. According previous articles [39,40], this drawback could be overcome to some extent by performing a second intrusion process. The difference between the first and second intrusion is the ink-bottle porosity, by deducting which the effective porosity could be obtained. It is this effective porosity that governs the transport of water and ions in cementitious system. The inclusion of GNP particles might increase the degree of ink-bottle effect since the 2D layers could narrow the pores, as illustrated in Fig. 5. The addition of GNP might result in more ink-bottle pores, thus leading to lower effective porosity compared to the plain cement mortar. Furthermore, as revealed by Ye [41], the critical pore size would not change in the first and second intrusion. Therefore, the determined critical pore size remains valid in this study.

dVp/dlogDp, mL/g Pore volume (Vp), mL/g

dVp/dlogDp, mL/g

Pore volume (Vp), mL/g

The transport properties of cementitious materials are closely related to their capillary pore structure. The pore size distribution and capillary porosity for cement mortar with GNP A3775 and SP were probed using MIP test and the results are displayed in Fig. 4 and Table 3. Cement mortar with the same amount of SP (50% by weight of GNP added) used for the dispersion of GNP was also tested to differentiate between the effects of SP and GNP. According to Mindess et al. [2], the capillary pore sizes can be divided into three categories, namely large capillaries or macropores (50–10,000 nm), medium capillaries or large mesopores (10–50 nm) and small isolated capillaries or small mesopores (2.5– 10 nm). Transport properties such as permeation and diffusion of cement paste are affected more by the macropores [2] and are studied in greater detail by comparing the fraction of macropores and the critical pore size. The cumulative pore volume curves are shown in Fig. 4(a) and (b) for cement mortar with GNP + SP and cement mortar with SP respectively. With the addition of GNP + SP, the cumulative pore volume curves in Fig. 4(a) shift downwards, indicating a refinement of the pore structure. This downward shift can be attributed to the GNP since the corresponding curves for cement mortar with SP only shows an upward shift in Fig. 4(b). The comparison between the fraction of macropores, as well as the average, median and critical pore diameters in Table 3 will further reveal the effects of GNP on the pore structure. The volume fraction of macropores was calculated from the area under the MIP curve within the macropore size (between 50 and 10,000 nm) over the total under the curve within the whole size range (3.7 nm–10 000 nm). With the addition of 1.25% SP by wt of cement, the fraction of macropores decreased marginally by 4% but further increase in SP of up to 3.75% resulted in 38% increase in the fraction of macropores. When GNP is also included, there is a marginal reduction of 3% when 2.50% GNP + 1.25% SP are added and this can be due to the effect of SP which has resulted in similar reduction. Further addition of GNP and SP of up to 7.5% and 3.75% respectively resulted in a 12% reduction in the fraction of macropores. This is consistent with the other trends in average, median and critical pore sizes, which shrink when both GNP and SP are added. The critical pore size, which corresponds to the peak in the differential distribution curves shown in Fig. 4(b) and (d), is tabulated in Table 3. Critical pore size represents the size of pore entryways that allows maximum percolation throughout the pore system [2] which has a

10

13

0.12 0.09

(b)

0SP (Ref.) 1.25SP 2.50SP 3.75SP

0.06 0.03 0.00 0.12

(d)

0.09 0.06 0.03 0.00 1E-3

0.01

0.1

1

10

Pore diameter (Dp), m

Fig. 4. Cumulative pore volume curves for mortar with (a) GNP and SP; and (b) SP only; and corresponding differential distribution curves for mortar with (c) GNP and SP; and (d) SP only.

14

H. Du, S.D. Pang / Cement and Concrete Research 76 (2015) 10–19

Table 3 MIP results for cement mortar with GNP A3775.

Reference GNP + SP

SP only

GNP content, % by wt of cement

SP content, % by wt of cement

Porosity, %

Average pore diameter, nm

Median pore diameter, nm

Critical pore diameter, nm

Fraction of macropores, %

0 2.50

0 1.25

5.00

2.50

7.50

3.75

0

1.25

0

2.50 3.75

20.0 16.9 (−15.5) 17.9 (−10.5) 18.0 (−10.0) 28.4 (+42.0) 42.1 (+111) 76.6 (+283)

62.1 54.6 (−12.1) 45.0 (−27.5) 44.1 (−29.0) 91.9 (+48.0) 1396 (+2150) 1489 (+2300)

110.7 73.9 (−33.2) 73.6 (−33.5) 73.9 (−33.2) 110.4 (−0.3) –a

0

22.2 20.2 (−9.0) 24.1 (+8.6) 20.3 (−8.6) 19.7 (−11.3) 24.7 (+11.3) 26.3 (+18.5)

54.3 52.7 (−3.0) 48.6 (−10.5) 47.9 (−11.8) 52.2 (−3.9) 66.8 (+23.0) 74.8 (+37.8)

–a

Values in parenthesis denote the percentage change when compared to reference. a Critical pore size is irrelevant since the entrained air pockets get too big with high SP content.

3.2. Compressive and flexural strength Before further investigation on the transport properties of cement mortar with GNP, the effect of GNP on the strength of the cement mortar is being investigated. The amount of GNP to be added should not adversely affect the mechanical strength of the cement mortar. The compressive and flexural strengths of cement mortar with GNP are tested at 28 days and the results are shown in Fig. 6. The effect of GNP on the compressive strength is evaluated statistically with a null hypothesis of GNP having no effect on the compressive strength. The null hypothesis is supported at a significance level of 5% and this can be explained by the small variation in the porosity [2]. The same can be concluded for the insignificant effect of GNP on the flexural strength but a decreasing trend of flexural strength with higher GNP content is obvious and this is supported by a decreasing p-value from 0.721 at 5.0% GNP to 0.148 at 7.5% GNP. Compared to GNP polymer nanocomposites which have been reported to have a higher flexural strength [19], a major difference for mortar is the existence of sand particles and the interface transition zone (ITZ) between cement paste and sand particles. The GNPs are larger than the calcium silicate hydrate (CS-H) particles and of the same length scale as the ITZ but are smaller than the sand particles. While the addition of GNP could have a strengthening effect on the cement paste due to its size that spans over several C-S-H particles and a high aspect ratio (as shown in Fig. 1(b)), it is unlikely to have an influence on the ITZ or the cement mortar matrix. The size of the GNP is insufficient to transfer the stresses over several sand particles to mitigate the weak zones in the microstructure or to provide adequate anchorage to bridge over the ITZ or the microcracks that exist between the sand particles. At much higher GNP content, both the compressive and the flexural strengths drop and it is likely due to two reasons. First, a larger amount

of SP is used to disperse a higher GNP content. The addition of SP only, while keeping the w/c ratio constant, will lead to the formation of excessive foam which increases the bleeding tendency and porosity. This will have an adverse effect on both the compressive and flexural strengths as shown in Fig. 6. With 3.75% SP, the compressive and flexural strengths drop by 56% and 45% respectively as compared to plain cement mortar without GNP. Second, the time taken for sonicating the GNP is kept the same for all GNP contents which may not be sufficient for effective dispersion at high GNP contents, the result of which will lead to GNP agglomerates in a weakened cement matrix (as shown in Fig. 1(c)). To date, there is no agreement on the influence of nano-particle with high aspect ratio, such as CNT, on the compressive and flexural strengths of cement composites [10,42,43]. 3.3. Water penetration depth The water penetration profiles for the specimens with different amount of GNP and SP are shown in Fig. 7. The average water penetration depths in cement mortar with GNP and SP are measured and computed as shown in Table 5. The effect of GNP is contrasted by carrying out the same test on cement mortar with SP only. With the addition of SP only, the cement mortar became very porous, resulting in much higher average water penetration depth as compared to the reference specimen shown in Table 5. When GNP is added with the corresponding amount of SP, the penetration depth drops by 75% from 18.2 mm for the reference cement mortar with no GNP to 4.4 mm for the cement mortar with 5% GNP + 2.5% SP. This improved resistance to water penetration

Table 4 Average sizes for different categories of pores from MIP tests.

Reference GNP + SP

SP only

GNP content, % by wt of cement

SP content, % by wt of cement

Fraction of air voids, %

Average macropore diameter, nm

Average large mesopore diameter, nm

Average small mesopore diameter, nm

0 2.50 5.00 7.50 0 0 0

0 1.25 2.50 3.75 1.25 2.50 3.75

8.58 14.2 11.1 7.52 6.41 40.8 52.3

115 117 84.7 81.4 115 160 185

18.8 19.0 19.5 19.2 22.5 20.5 22.9

5.48 5.31 5.31 5.53 5.37 5.33 5.34

Fig. 5. Illustration of the addition of GNP on the mercury intrusion into cementitious system, adopted from Ref. [41].

Compressive strength, MPa

H. Du, S.D. Pang / Cement and Concrete Research 76 (2015) 10–19

15

70

(a) 60 50 40 30 20 SP+GNP SP

10 0

0.0

2.5

5.0

7.5

GNP content, %

Flexural strength, MPa

10

(b)

8 6 4 2

SP+GNP SP

Fig. 7. Results of water permeability tests for cement mortar with (a) GNP + SP and (b) SP only.

0 0.0

2.5

5.0

7.5

GNP content, % Fig. 6. Effect of GNP and SP on the flexural strength of cement mortar.

can be attributed to the reduced critical pore diameter due to the refinement of the pore structure by the GNP, as well as extensive barriers formed by the impermeable GNP in the cement matrix which results in more tortuous path for the ingress of water as shown in Fig. 1(b). The resistance to water penetration of the cement mortar increases with the addition of more GNP but at much higher GNP content of 7.5%, the resistance drops as compared to cement mortar with 5.0% GNP. This is likely due to the non-uniform dispersion of GNP in the cement paste at this high concentration, causing clustering of the GNP which compromises its blocking efficiency to the ingress of water. The effect of clustering was also consistent with the drop in flexural strength that is reported in the previous section when much higher GNP content is used. Ji [14] reported a 45.5% reduction in average water penetration depth in concrete by adding 3.6% nano-SiO2 which has a nano-filler effect. Compared with the spherical shape of nano-SiO2, GNP can achieve better effectiveness with a reduction of 64% in average water penetration depth in cement mortar by using only 2.5% GNP. This is due to its layered morphology which helps in the refinement of the pore structure and the tortuosity of the path for water ingress. While the tests in this paper were based on cement mortar, similar effectiveness is expected if GNP is introduced into concrete.

diffusion process of chloride can be predicted by Crank's solution to Fick's second law [2], as follows:    x C ðx;t Þ ¼ C s 1−erf pffiffiffiffiffiffiffiffi 2 Dc t

where C(x,t) is the measured chloride concentration at depth x and exposure time t, Cs is the chloride concentration in the specimen at the surface, Dc is to the apparent chloride diffusion coefficient, and erf is the error function. Dc and Cs are determined by curve fitting the expression in Eq. (1) to the spatial variation of chloride concentration shown in Fig. 8 and the values are tabulated in Table 6. Cement mortar with 2.5% GNP has the lowest apparent chloride diffusion coefficient, which is only 30% of the value for plain cement mortar; on the other hand, it suffers from a slightly higher chloride concentration at the surface Cs as compared to plain cement mortar and that is likely due to the low concentration of the GNP which is ineffective to act as a barrier against chloride ion

Table 5 Average water penetration depths for cement mortar with GNP A3775.

Reference GNP + SP

3.4. Chloride diffusion SP only

Fig. 8 displays the chloride concentration profiles for mortar with various GNP contents after 28 days of immersion in NaCl solution. The

ð1Þ

GNP content, % by wt of cement

SP content, % by wt of cement

Average water penetration depth, mm

Change as compared to reference, %

0 2.50 5.00 7.50 0 0 0

0 1.25 2.50 3.75 1.25 2.50 3.75

18.1 ± 1.31 6.51 ± 2.56 4.39 ± 0.64 9.00 ± 0.50 24.6 ± 1.28 95.4 ± 11.3 142.4 ± 5.2

– −64.1 −75.8 −50.4 +35.4 +426 +685

H. Du, S.D. Pang / Cement and Concrete Research 76 (2015) 10–19 Table 7 Changes in chloride penetration with time for cement mortar with GNP A3775.

0.7 0GNP+0SP (Ref.) 2.5GNP+1.25SP 5.0GNP+2.5SP 7.5GNP+3.75SP

0.6 0.5

GNP content, % by SP content, % by Chloride penetration wt of cement wt of cement depth, mm

0.4

Reference

0 2.50 GNP + SP 5.00 7.50

0.3

0.1 0.0 0

10

20

30

40

50

Distance, mm Fig. 8. Chloride content profiles and best-fitted curves for cement mortar with GNP.

penetration at the surface. The apparent chloride diffusion coefficients for cement mortar with 5.0% and 7.5% of GNP are both reduced by at least 40% in comparison with plain cement mortar and their chloride concentration at the surfaces are decreased by 25% and 39% respectively. The reduction of both Dc and Cs for cement mortar with 5.0% and 7.5% GNP is depicted in Fig. 8 where the curves for the spatial distribution of chloride concentration of these two GNP contents are shifted downwards. The effect of these two reductions can potentially have an effect of increasing the durability of the cement mortar against corrosion by 2–3 times. Table 7 shows the chloride penetration depth in cement mortar with various contents of GNP after different periods of immersion in NaCl solution, namely 3 days, 7 days, 14 days and 28 days. With the passage of time, the chloride ion penetration increases for all specimens with plain cement mortar suffering from the highest chloride penetration at all ages tested while cement mortar with GNP offers better resistance against chloride penetration. For example, the chloride penetration depth after immersion for 14 days was reduced by 31%, 48% and 37% by adding 2.5%, 5.0% and 7.5% of GNP respectively. The relationship between the chloride penetration depth xf and the time of measurement pffiffi can be easily obtained by rearranging Eq. (1) which leads to x f ¼ k t where   pffiffiffiffiffiffi  C −1 1− f k ¼ Dc  2  erf Cs

ð2Þ

pffiffi The depth of chloride penetration xf is plotted against t and the linearity of the relationship is verified by the curves in Fig. 9. The apparent diffusion coefficient Dc is proportional to the square of the slope of the curves k2 and using this as an estimate of Dc, we again expect a reduction of approximately 50% with the addition of GNP which is consistent with the results in Table 6. The addition of as little as 2.5% GNP can result in significant improvement in the chloride diffusion resistance of the cement mortar. This is due to the reduction in critical pore diameter (as shown in Table 4) Table 6 Chloride diffusion coefficients for cement mortar with GNP A3775.

Reference GNP + SP

0 1.25 2.50 3.75

t=3 days

t=7 days

t = 14 days

t = 28 days

5.54 4.60 3.68 3.85

8.07 5.84 6.43 5.94

12.74 8.85 6.67 8.05

20.00 14.08 16.30 13.97

0.2

GNP content, % by wt of cement

SP content, % by wt of cement

Dc, ×10−11 m2/s

C s, %

R2

0 2.50 5.00 7.50

0 1.25 2.50 3.75

7.02 2.15 4.10 3.91

0.61 0.70 0.46 0.37

0.998 0.990 0.999 0.983

when GNP is added to refine the pore structures. The addition of the impermeable GNP will also lead to larger tortuosity which contributes to the enhanced barrier property of the cement mortar against chloride diffusion. The refinement of the pore structure by the GNP has led to consistent effect on the water permeability resistance and chloride diffusion resistance of the cement mortar. It should be noted that the addition of GNP beyond 2.5% does not dramatically affect the resistance as the critical pore diameter is not much affected. Compared with the effectiveness of other nano-particles reported by Zhang and Li [16], the addition of 3% nano-SiO2 and 3% nano-TiO2 resulted in 10.5% and 19.6% reduction in the chloride diffusion respectively but a much higher reduction of 69.7% can be obtained by adding 2.50% GNP. 3.5. Chloride migration The non-steady-state migration coefficient Dnssm for the rapid chloride migration (RCM) tests on mortar with GNP at 28 days can be determined using the following expression [30] with the units for L and xd in mm and Dnssm in 10−12 m2/s: Dnssm ¼

0:0239ð273 þ T ÞL xd −0:0238 ðU−2Þt

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! ð273 þ T ÞLxd U−2

ð3Þ

The parameters U = 20 V and t = 6 hours are kept constant for all the specimens and the measured average temperature for all tests is T = 29 °C. We can simplify Eq. (3) into the following equation with the measured values of L, xd given in Table 8: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

Dnssm ¼ 5:97  10−4 L xd −0:0238 16:78Lxd

ð4Þ

With as little as 2.5% GNP + 1.25% SP, the migration coefficient drops by 31% from 19.8 × 10−12 m2/s to 13.6 × 10−12 m2/s and the lowest

Chloride penetration depth, mm

Chloride content, % wt. of mortar

16

20

15

0GNP+0SP (Ref.) 2.5GNP+1.25SP 5.0GNP+2.50SP 7.5GNP+3.75SP

10

5

0

Fig. 9. Change in chloride diffusion depth with the square root of time.

H. Du, S.D. Pang / Cement and Concrete Research 76 (2015) 10–19

of tortuosity in reducing the transport of water can be evaluated with the tortuosity factor τp [46] which is expressed as follows:

Table 8 Chloride migration coefficients for cement mortar with GNP A3775.

Reference GNP + SP

SP only

GNP content, % by wt of cement

SP content, % by wt of cement

Lavg, mm

xd,avg, mm

Dnssm, ×10−12 m2/s

Change as compared to reference, %

0 2.50 5.00 7.50 0 0 0

0 1.25 2.50 3.75 1.25 2.50 3.75

52.0 50.9 51.2 53.8 50.8 52.5 51.8

7.64 5.64 5.32 5.41 7.01 15.4 28.9

19.8 13.6 12.8 13.5 17.6 44.4 86.9

– −31.4 −35.3 −31.8 −11.3 +125 +339

migration coefficient of 12.8 × 10−12 m2/s achieved with 5.0% GNP + 2.50% SP as shown in Table 8. Addition of GNP and SP beyond 2.50% and 1.25% respectively does not result in significant reduction in the migration coefficient. The influence of GNP on chloride migration is consistent with the resistance to chloride ion diffusion. Further tests were conducted on cement mortar with SP only which led to severe vulnerability of the cement mortar to resist chloride migration as shown in Table 8. With the addition of 1.25% SP which corresponds to the amount added for the dispersion of 2.5% GNP, there is a slight reduction in the migration coefficient. Further increase in the SP to 2.50% and 3.75% resulted in a highly porous mix that led to multifold increases in the migration coefficients. It is evident from the comparison that the improvement in the chloride migration resistance can be attributed solely to the effect of GNP. The foam accompanying SP especially at high dosage entrains more voids in the cement mortar, leading to higher porosity and larger average pore size. Cement mortar with only SP exhibits the largest average pore size and at the same time has the lowest resistance to chloride migration when the dosage of SP is the highest. Also, bleeding was observed for this mixture, which could result in continuous flow paths in the matrix and the weak zone trapped beneath the sand particles. The relationship between migration coefficient and pore structure (in terms of average pore diameter) is established as shown in Fig. 10. With larger pore size, the chloride migration coefficient increases linearly. This observed correlation agrees well with those previously reported [44,45]. 3.6. Mechanism of transport barrier by GNP The introduction of the impermeable GNP in the mortar will lead to more tortuous paths for the ingress of water and chloride ions. Considering an ideal case where all the 2-D particles are uniformly dispersed along the preferred orientation shown in Fig. 11(a), the contribution

-12

2

Migration coefficient, x10 m /s

17

100 80

λ τp ¼ 1 þ ϕ 2

ð5Þ

where λ and ϕ are the aspect ratio and the volume fraction of GNP in the matrix, respectively. This factor is an upper bound estimate of the reduction in transport due to tortuosity since uniform dispersion and perfect alignment of platelets have been assumed. The assumption of perfect alignment can be relaxed by introducing an orientation parameter S which accounts for the orientation of the platelets [47] and the modified tortuosity factor τr is given by:   λ 2S þ 1 τr ¼ 1 þ ϕ 2 3

ð6Þ

When all the platelets are perfectly aligned normal to the direction of flow shown in Fig. 11(a), S takes the value of 1 and when the platelets are randomly oriented as shown in Fig. 11(b), S takes the value of 0. Water permeability coefficient (KW) could be estimated by equation developed by Valenta [48], which is based on the water penetration depth (d) obtained.

2

KW ¼

d v 2ht

ð7Þ

where v is the porosity of cement system, h is the water pressure applied to the specimen, and t is the testing time. According to Nielsen [46], the permeability of particle-filled (KW′) and unfilled composites (KW) could be expressed as K 0W ¼ K W =τ:

ð8Þ

This theoretical estimate is compared with the experimentally measured permeability ratio K′W/KW shown in Fig. 12(a). The further reduction in water permeability beyond the effect of tortuosity can be attributed to the refinement in pore structure brought about by the GNP. The tortuosity effect contributes about 20% and 32% reduction in water permeability coefficient for mortar with 0.6 vol.% and 1.2 vol.% of GNP respectively. However, at higher GNP content of 1.8 vol.%, the water penetration depth was expected to fall but instead rose to 0.2 times that of plain mortar. Since the fineness of the pore structure remains almost the same as the mortar with 1.2 vol.% GNP, it is likely that the clustering of the GNP shown in Fig. 1(c) has resulted in its lower efficiency in improving the barrier properties of the mortar. The ingress of chloride ions by diffusion is described by the solution to Fick's second law of diffusion in Eq. (1). If the distance x traveled by the chloride ions from the exposed surface takes into account the length of the tortuous path such that x′ = τx, it results in an apparent chloride diffusion coefficient D′ which can be expressed as

60

D0c ¼ Dc =τ 2

40

This modification of Dc is based on tortuosity argument only and if we plot Dc′/Dc, we can evaluate the relative contribution of tortuosity in the chloride diffusion barrier. From the comparison of Dc′/Dc against 1/τ 2, as shown in Fig. 12(b), we can expect tortuosity to contribute almost fully to the choloride diffusion barrier for GNP contents of 1.2 vol.% and above. At GNP content of 0.6 vol.%, there is a high tortuosity contribution of 60% against the chloride diffusion barrier. At high GNP content of 1.8 vol.%, the Dc′/Dc ratio is significantly less than the contribution expected from tortuosity and this is likely due to the agglomeration of GNP. From inverse calculation, the effective aspect ratio at 1.8 vol.% of GNP is estimated to be half of the value given in Table 2.

20

GNP+SP SP Regression line

0

Average pore diameter, nm Fig. 10. Relationship between migration coefficient Dnssm and average pore diameter.

ð8Þ

18

H. Du, S.D. Pang / Cement and Concrete Research 76 (2015) 10–19

Fig. 11. Effect of platelet orientation on the tortuous path: (a) platelets are perfectly aligned normal to the direction of flow, and (b) platelets are randomly oriented.

The non-steady state migration coefficient of chloride ions under an applied potential difference can be described by Eq. (3). Dividing Eq. (3) throughout by L2/t for the dimensionless form, we obtain Dnssm t L2

0:0239ð273 þ T Þ xd −0:0238 ¼ L ðU−2Þ

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi! ð273 þ T Þ xd U−2 L

ð9Þ

from the experiments. This is again likely due to the agglomeration of the GNP as the sonication time is kept constant for all GNP contents. It should be noted that due to the inert properties of graphene, the chloride ions were impeded by the GNPs due to the impermeable barrier effect instead of being absorbed [49]. 4. Conclusions

Similar to the treatment of chloride diffusion coefficient, we can evaluate the contribution of tortuosity to the non-steady state migration coefficient by comparing Dnssm′/Dnssm against 1/τ 2, as displayed in Fig. 12(b). Assuming that the nanoplatelets are uniformly dispersed, tortuosity can account for all the changes in Dnssm at 0.6 vol.% GNP. At much higher GNP content of 1.8 vol.%, the predictions for Dnssm′/Dnssm based on tortuosity arguments over-estimated the value obtained

1.0

(a)

0.6

'

K w/Kw

0.8

0.4 0.2 0.0

1.0

(b)

Acknowledgments The financial support from A*STAR SERC (grant number R-302-000034-305) is acknowledged. The authors also wish to express their thanks to Asbury Graphite Mills for supplying GNP materials.

0.8 0.6

References

'

D /D

This study demonstrates that the barrier properties of cement mortar can be improved by adding low cost GNP through a facile fabrication, which refines the pore structure of the cement mortar, as revealed by the MIP, due to its large aspect ratio and layered structure, as well as its barrier characteristics. The resistance of the cement mortar to water permeability, chloride diffusion and chloride migration are increased by 30–70% with the addition of GNP which are higher than that reported in literature for cement composites containing nanosized spherical particles such as nano-SiO2 or nano-TiO2. The effectiveness of the GNP in enhancing the barrier properties of the cement mortar is demonstrated by the addition of as little as 2.5% GNP which reduces the critical pore diameter by more than 30% but further addition of GNP does not lead to dramatic improvement. At 2.5% GNP and below, the tortuous path introduced contributes up to 20% of the barrier effects for water ingress and attributed almost fully to the barrier effects for chloride ion ingress. At 7.5% GNP and above, the barrier properties may suffer due to lowered dispersion efficiency which resulted in nano-particle agglomeration. The extent of agglomeration can be represented by the effective aspect ratio that is estimated using the tortuosity model and is found to be reduced by 50–60% at 7.5% GNP. Superplasticizer provides no additional contribution to the improved durability while in fact compromises it at high dosages. Given its abundance, low cost, facile processing and enhanced barrier properties in cementitious composites, GNP has excellent potential for use in large scale for building materials.

0.4 0.2 0.0

Diffusion coefficient Migration coefficient

GNP volume fraction, % Fig. 12. Comparison of predictions in reduction of transport properties based on tortuosity arguments (τp from Eq. (5) and τr from Eq. (6)) and experimentally measured reduction in (a) water permeability coefficient, and (b) chloride diffusion and migration coefficient.

[1] A.M. Neville, Properties of Concrete, fourth ed. Prentice Hall, London, 1995. [2] S. Mindess, J.F. Young, D. Darwin, Concrete, second ed. Prentice Hall, New Jersey, 2003. [3] P.K. Mehta, P.J.M. Monteiro, Concrete Microstructure, Properties, and Materials, third ed. McGraw-Hill, New York, 2006. [4] P.K. Mehta, Reducing the environmental impact of concrete, Concr. Int. (10) (2001) 61–66. [5] M. Sahmaran, M. Li, V.C. Li, Transport properties of engineered cementitious composites under chloride exposure, ACI Mater. J. 104 (2007) 604–611. [6] P.C. Aitcin, The durability characteristics of high performance concrete: a review, Cem. Concr. Compos. 25 (2005) 409–420. [7] S.P. Shah, S.H. Ahmad, High Performance Concrete: Properties and Applications, McGraw-Hill, New York, 1994. [8] M.H. Zhang, V.M. Malhotra, High-performance concrete incorporating rice husk ash as a supplementary cementing material, ACI Mater. J. 93 (1996) 629–636.

H. Du, S.D. Pang / Cement and Concrete Research 76 (2015) 10–19 [9] K. Sobolev, M. Ferrada-Gutierrez, How nanotechnology can change the concrete world: part I, Am. Ceram. Soc. Bull. 84 (2005) 14–17. [10] Z.S. Metaxa, J.W.T. Seo, M.S. Konsta-Gdoutos, M.C. Hersam, S.P. Shah, Highly concentrated carbon nanotube admixture for nano-fiber reinforced cementitious materials, Cem. Concr. Compos. 34 (2012) 612–615. [11] J. Chen, S.C. Kou, C.S. Poon, Hydration and properties of nano-TiO2 blended cement composites, Cem. Concr. Compos. 34 (2012) 642–649. [12] S. Kawashima, P. Hou, D.J. Corr, S.P. Shah, Modification of cement-based materials with nanoparticles, Cem. Concr. Compos. 36 (2013) 8–15. [13] G.Y. Li, P.W. Wang, X. Zhao, Pressure-sensitive properties and microstructure of carbon nanotube reinforced cement composites, Cem. Concr. Compos. 29 (2007) 377–382. [14] T. Ji, Preliminary study on the water permeability and microstructure of concrete incorporating nano-SiO2, Cem. Concr. Res. 35 (2005) 1943–1947. [15] J.J. Gaitero, I. Campillo, A. Guerrero, Reduction of the calcium leaching rate of cement paste by addition of silica nanoparticles, Cem. Concr. Res. 38 (2008) 1112–1118. [16] M.H. Zhang, H. Li, Pore structure and chloride permeability of concrete containing nano-particles for pavement, Constr. Build. Mater. 25 (2011) 608–616. [17] D. Kong, X. Du, S. Wei, H. Zhang, Y. Yang, S.P. Shah, Influence of nano-silica agglomeration on microstructure and properties of the hardened cement-based materials, Constr. Build. Mater. 37 (2012) 707–715. [18] F. Sanchez, K. Sobolev, Nanotechnology in concrete — a review, Constr. Build. Mater. 24 (2010) 2060–2071. [19] B.Z. Jang, A. Zhamu, Processing of nanographene platelet (NGPs) and NGP nanocomposites: a review, J. Mater. Sci. 43 (2008) 5092–5101. [20] O.C. Compton, S. Kim, C. Pierre, J.M. Torkelson, S.T. Nguyen, Crumpled graphene nanosheets as highly effective barrier property enhancers, Adv. Mater. 22 (2010) 4759–4763. [21] J. Lu, I. Do, L.T. Drzal, R.M. Worden, I. Lee, Nanometal-decorated exfoliated graphite nanoplatelet based glucose biosensors with high sensitivity and fast response, ACS Nano 2 (2008) 1825–1832. [22] K. Kostarelos, K.S. Novoselov, Exploring the interface of graphene and biology, Science 344 (2014) 261–263. [23] H. Du, S.D. Pang, S.T. Quek, Smart multifunctional cement mortar containing graphite nanoplatelet, in: J.P. Lynch, C.B. Yun, K.W. Wang (Eds.),SPIE, vol. 8692, SPIE Press, San Diego, USA 2013, p. 869223, http://dx.doi.org/10.1117/12.2009005. [24] S.D. Pang, H.J. Gao, C. Xu, S.T. Quek, H. Du, Strain and damage self-sensing cement composites with conductive graphene nanoplatelet, in: J.P. Lynch, K.W. Wang, H. Sohn (Eds.),SPIE, vol. 9061, SPIE Press, San Diego, USA 2014, p. 9061126, http:// dx.doi.org/10.1117/12.2045329. [25] J.-L. Le, H. Du, S.D. Pang, Use of 2-D graphene nanoplatelet (GNP) in cement composites for structural health evaluation, Compos. Part B 67 (2014) 555–563. [26] ASTM C 109, Standard test method for compressive strength of hydraulic cement mortars (using 2-in. [50 mm] cube specimens), ASTM International, West Conshohocken, PA, 2005. [27] A. Ciesielski, P. Samori, Graphene via sonication assisted liquid-phase exfoliation, Chem. Soc. Rev. 43 (2014) 381–398. [28] Z.Y. Xia, S. Pezzini, E. Treossi, G. Giambastiani, F. Corticelli, V. Morandi, A. Zanelli, V. Bellani, V. Palermo, The exfoliation of graphene in liquids by electrochemical, chemical, and sonication-assisted techniques; a nanoscale study, Adv. Funct. Mater. 23 (2013) 4684–4693. [29] ASTM C 348, Standard test method for flexural strength of hydraulic cement mortar, ASTM International, West Conshohocken, PA, 2002.

19

[30] BS EN 12390-8, Testing method concrete part 8: depth of penetration of water under pressure, British Standard, 2009. [31] NT BUILD 443, Concrete, hardened: accelerated chloride penetration, NORDTEST, 1995. [32] BS 1881-124, Testing concrete—part 124: method for analysis of hardened concrete, British Standard, 1988. [33] NT BUILD 492, Concrete, mortar and cement-based repair materials: chloride migration coefficient from non-steady-state migration experiments, NORDTEST, 1995. [34] G.Y. Li, P.M. Wang, X. Zhao, Mechanical behavior and microstructure of cement composites incorporating surface-treated multi-walled carbon nanotubes, Carbon 43 (2005) 1239–1245. [35] S.J. Chen, F.G. Collins, A.J.N. Macleod, Z. Pan, W.H. Duan, C.M. Wang, Carbon nanotube-cement composites: a retrospect, IES J. A Civ. Struct. Eng. 4 (2011) 254–265. [36] T. Nochaiva, A. Chaipanich, Behavior of multi-walled carbon nanotubes on the porosity and microstructure of cement-based materials, Appl. Surf. Sci. 257 (2011) 1941–1945. [37] J.M. Makar, G.W. Chan, Growth of cement hydration products on single-walled carbon nanotubes, J. Am. Ceram. Soc. 92 (2009) 1303–1310. [38] S. Diamond, Mercury porosimetry: an inappropriate method for the measurement of pore size distributions in cement-based materials, Cem. Concr. Res. 30 (2000) 1517–1525. [39] J. Kaufmann, R. Loser, A. Leemann, Analysis of cement-bonded materials by multicycle mercury intrusion and nitrogen sorption, J. Colloid Interface Sci. 336 (2009) 730–737. [40] J. Zhou, G. Ye, K. van Breugel, Characterization of pore structure in cement-based materials using pressurization-depressurization cycling mercury intrusion porosimetry (PDC-MIP), Cem. Concr. Res. 40 (2010) 1120–1128. [41] G. Ye, Experimental study and numerical simulation of the development of the microstructure and permeability of cementitious materials(PhD Thesis) Delft University of Technology, 2003. [42] M.S. Konsta-Gdoutos, Z.S. Metaxa, S.P. Shah, Highly dispersed carbon nanotube reinforced cement based materials, Cem. Concr. Res. 40 (2010) 1052–1059. [43] F. Collins, J. Lambert, W.H. Duan, The influences of admixtures on the dispersion, workability, and strength of carbon nanotube–OPC paste mixtures, Cem. Concr. Compos. 34 (2012) 201–2017. [44] H.Y. Moon, H.S. Kim, D.S. Choi, Relationship between average pore diameter and chloride diffusivity in various concretes, Constr. Build. Mater. 20 (2006) 725–732. [45] C.C. Yang, On the relationship between pore structure and chloride diffusivity from accelerated chloride migration test in cement-based materials, Cem. Concr. Res. 36 (2006) 1304–1311. [46] L.E. Nielsen, Models for the permeability of filled polymer systems, J. Macromol. Sci. Chem. A 1 (1967) 929–942. [47] R.K. Bharadwaj, Modelling the barrier properties of polymer-layered silicate nanocomposites, Macromolecules 34 (2001) 9189–9192. [48] O. Valenta, The permeability and durability of concrete in aggressive conditions, Proceedings of 10th International Congress on Large Dam, Monteral 1970, pp. 103–117. [49] M.V. Fedorov, R.M. Lynden-Bell, Probing the neutral graphene-ionic liquid interface: insights from molecular dynamics simulations, Phys. Chem. Chem. Phys. 14 (2012) 2552–2556.