Copyright WILEY-VCH Verlag GmbH & Co. KGaA, 69469 Weinheim, Germany, 2018.

Supporting Information
for Adv. Funct. Mater., DOI: 10.1002/adfm.201705183

Ultrahigh Performance Nanoengineered Graphene–Concrete

Composites for Multifunctional Applications
Dimitar Dimov, Iddo Amit, Olivier Gorrie, Matthew D.
Barnes, Nicola J. Townsend, Ana I. S. Neves, Freddie Withers,
Saverio Russo, and Monica Felicia Craciun*
 Supplementary Materials
Ultrahigh Performance Nanoengineered Graphene-
Concrete Composites for Multifunctional Applications
Dimitar Dimov, Iddo Amit, Olivier Gorrie, Ana I. S. Neves, Nicola J. Townsend,
Matthew D. Barnes, Freddie Withers, Saverio Russo, Monica F. Craciun*

correspondence to: m.f.craciun@exeter.ac.uk

S1. Liquid exfoliation of Graphene

Different initial amounts of graphite flakes and sodium cholate per volume of
water were blended using L5M high-shear mixer to investigate the resulting
graphene content on the various suspensions used for reinforcing concrete.
Table S1 summarizes the study on graphene suspensions. We have
investigated various initial amounts of starting graphite, sodium cholate and water,
including those reported in the literature. The first initial amounts were 50mg/mL
graphite and 10mg/mL sodium cholate and were based on Paton et al.’s work, [1].
However we found that this results in a large amount of graphite residues left after
the shear exfoliation. Therefore we developed suspensions where the initial amount
(i.e. graphite and sodium cholate to water ratio) was progressively reduced but the
graphite to sodium cholate ratio was maintained at 5:1. The initial weight of the
membrane before and after filtration was compared to measure the amount of
graphene per volume filtrated, here referred to as final concentration. The blending
parameters for the data reported in Table S1 are the optimized values which resulted
from a systematic Raman study of the effect of blending time and speed on the
resulting graphene dispersions, see Figures S1.

Input materials Blending Final

parameters concentration
Graphite/ Sodium cholate/ Graphite / Blending Blending Few layer
water water sodium time speed graphene / water
ratio ratio cholate ratio
50 g/L 10 g/L 5:1 2h 5000 rpm 2.4 g/L
25 g/L 5 g/L 5:1 2h 5000 rpm 0.9 g/L
12.5 g/L 2.5 g/L 5:1 2h 5000 rpm 0.6 g/L
6.25 g/L 1.25 g/L 5:1 2h 5000 rpm 0.3 g/L
Table S1. Summary of initial input materials, blending time and speed, and final
concentration of graphene flakes.

Figure S1a shows Raman spectra for graphene dispersions where the
blending speed was varied from 4000 rpm to 7000 rpm for different blending times
ranging from 30 min to 3h. To perform these measurements the graphene
dispersions were drop-casted on 1cm x 1cm SiO2/Si substrates and dried on a hot
plate at 100oC. Raman measurements were performed in various places on the
substrate and the representative spectra were plotted in Figure S1a. For all the
investigated blending parameters we observed that few-layer graphene flakes were
obtained. This is evident from the presence of the G and 2D peaks together with the
shape of the 2D peak, its relative intensity to the G peak, as well as the absence of a
doublet in the D and 2D peaks. Lorentzian fitting of the D-, G- and 2D-peaks allows
us to get an initial understanding on the structural ordering within the films and the
number of layers by analyzing the band intensities (ID,G,2D). By doing this analysis we
concluded that blending at 5000rpm for 2h are the parameters that give the lowest
level of defects (i.e. lowest ID/IG) with reasonable thickness of the few-layer (i.e.
I2D/IG =0.76). Figure S1b illustrates the dependence of I2D/IG and ID/IG ratios on the
blending time at 5000rpm blending speed. It is apparent that 2h of blending results in
graphene flakes with the lowest level of defects (i.e. the lowest I D/IG ratio). On the
other hand, Figure S1c shows the dependence of I2D/IG and ID/IG ratios on the
blending speed for fixed blending time of 2h. Again it is apparent that 5000 rpm
blending speed results in the lowest level of defects.

b c

Figure S1. a) Raman spectra of graphene dispersions obtained after various blending
times and speeds. b) Dependence of I2D/IG and ID/IG ratios on the blending time at
5000rpm fixed blending speed. c) dependence of I2D/IG and ID/IG ratios on the blending
speed for fixed blending time of 2h
 The influence of an intermediate centrifugation step in this method, as
reported by Paton et al., [1] was also investigated. We found that, upon 1h
centrifugation at 6000 rpm, the graphene content is reduced by a factor of 100. On
the other hand, centrifuged suspensions will result in films with a smaller number of
graphene layers, rather than few-layer-graphene composed of 4-11 layers. Although
these suspensions might be more suitable than uncentrifuged suspensions for other
applications, such small graphene content will not be enough to reinforce concrete.
For this reason the graphene suspensions used to produce the concrete cubes were
simply decanted to discard excess of unexfoliated graphite, and not centrifuged.
In an effort to reduce the price of the materials used in the reinforcement of
concrete with graphene, we have also investigated the possibility to use a lower cost
surfactant. Recently, it was shown that graphene could be produced using
household detergent to stabilise the exfoliated graphene [3]. The surfactant used in
this work was Fairy washing-up liquid (commonly known as Fairy Liquid, FL), a
common household dishwashing liquid, with a composition of 15–30% anionic
surfactants, 5–15% nonionic surfactants (specified by the manufacturer, Proctor &
Gamble). However these studies were performed using turbulence-assisted shear
exfoliation produced by high speed rotating blades. Therefore, we have investigated
the use of high-shear rotor–stator mixer to exfoliate graphene in Fairy fluid as an
alternative to sodium cholate. However, the suspensions produced with this
surfactant were found to be less stable than with sodium cholate. This might be due
to the fact that it is harder to control the concentration of Fairy than that of sodium
cholate, which caused some excess detergent to eventually crystallise and separate
from the suspension. Furthermore, we have also observed what seemed to be a
degradation over time of the suspensions produced with Fairy, which did not happen
with sodium cholate, see Figure S2.

Figure S2: Stability of graphene suspensions exfoliated using sodium cholate for a period
of 6 months, and degradation after 10 days of graphene suspensions exfoliated using
household detergent (Fairy).
S2. Transfer method from the liquid to the substrate

In order to characterize the surfactant-stabilized graphene dispersions, we

employed the method developed by Withers et al [4] to transfer graphene flakes
dispersed in a liquid onto SiO2 substrate. 100 ml of solution was filtered through a
0.025 μm mixed cellulose hydrophilic Millipore® membrane using vacuum filtration
apparatus. Then, the membrane was submerged in deionized water as shown on
Figure S3b and ‘scooped out’ with the SiO2 substrate.

Figure S3. Scheme of deposition process: a) Photograph of deposited film on

membrane. B) Detaching the graphene film from the membrane in deionised water. C)
The film floats on the surface of the water. D) Scooping out part of the film on SiO2
substrate.

S3. SEM analysis

SEM micrographs were taken to study the area of the flakes deposited on
SiO2 substrate. A specifically designed image analysis procedure scanned the
highlighted green flakes on Figure S4b and counted the number of pixels they
occupy. Using the 4 μm scale the area of one pixel was calculated. As an example,
the smallest area size was found on Fig. S4b and labeled ‘1’ – it has 1780 pixels and
area of 0.07 μm2, medium size ‘2’ has 21920 pixels with area of 0.86 μm2 and lastly –
the largest flake has a size of 1.06 μm2.
 Figure S4. A) SEM image of FG film on substrate; b) Highlighted flakes and estimating
the smallest and largest sizes.

S4. Estimation of number of layers in graphene films obtained

from dispersions from the analysis of the 2D Raman peak
In order to estimate the number of layers in both FG and IG graphene dispersions,
we followed the method described in [1]. First, the position of the 2D peak of bulk
graphite is determined and denoted as ωp then the shoulder on the left hand side of
the peak is estimated to be 30 cm-1 away and denoted as ωs. The corresponding
intensities for both peaks are identified from the data and used in the ratio: I g-ite = I ωp
/ I ωs. The same technique is used for the 2D peak of the graphene spectra and the
same ratio is defined: Ig-ene = I ωp / I ωs. The following relationship between the two is
defined as M:
𝐼𝑔−𝑒𝑛𝑒
𝑀=
𝐼𝑔−𝑖𝑡𝑒

Which gives a numerical value usually found to be <1. Finally, M is substituted in the
empirical equation proposed by3:
2
𝑁𝑔 = 100.84𝑀+0.45𝑀

Where Ng is the calculated number of layers per suspension.

S5. Raman map

Following the analysis of the 2D peak to determine the number of layers, a

Raman map of the various positions was plotted as shown on Fig. S5. This is the
same map as Fig. 2f from the main paper, but with different scale. It is shown that all
of the graphene flakes have shifted 2D position when compared to that of graphite
(2724 cm-1) used for exfoliation.
Figure S5. Raman map of the intensity of the 2D peak.

S6. Industrial Graphene characterisation

For our experimental group labeled ‘IG’ we have used graphene nanoplatelets,
purchased from Cheap Tubes, with the following characteristics: X & Y dimensions
> 2μm, average thickness of 8-15nm and specific surface area of: 500 -700 m2/g.
Figure S6 shows a Transmission Electron Microscope (TEM) image of the films
provided, reproduced with permission from the supplier characterization database.
This shows that the flakes consist of few layer graphene.

Figure S6. TEM image of IG flakes, reproduced with permission from [4].
 We have further characterized the material by Raman spectroscopy, shown in
Figure S7. For these measurements we followed the same procedure as the one
shown in Figure S3.

Fig. S7. A) Photograph of membrane with 0.025 μm filter with deposited IG graphene
dispersions; b) Optical image of 20ml of filtrated solution, detached from the filter in a)
before depositing it on SiO2 substrate; c) SiO2 substrate with the film from b); d) Single
Raman spectra of the IG graphene; e) 2D peak comparison with bulk graphite to
estimate the number of layers.

We can explain the efficient dispersion of IG in water without the need of

surfactant as follows. According to the manufacturer, the nanoplatelets are
“chemically exfoliated from natural graphite”. This means they have been obtained
by oxidation of graphite, exfoliation, and subsequent reduction to reduced graphene
oxide (rGO). Unlike FG, which is obtained from graphite without any chemical
treatment, IG has a residual oxygen content, particularly towards the edges of the
platelets [6]. Although the manufacturer does not disclose the percentage of oxygen,
it does say that the graphene is 97% pure. We believe that the presence of the
residual oxygen-containing groups from the chemical process in which the graphene
nanoplatelets were produced makes IG much more hydrophilic than FG. Moreover,
the difference in size of the platelets of both types of graphene can also explain the
difference in dispersibility. FG samples are micrometers in lateral size (as indicated
in Fig S4), whereas the IG samples are bigger, within around 2um in area, according
to the manufacturer’s specifications [4]. As a simple guide, dispersing IG follows the
trend GO > rGO > G [6,7].

S7. Concrete preparation and results

The concrete samples were prepared and tested according to two standards,
recognized and used by industry. The first one – BS 1881-108:1983 certifies the
sample preparation, including dimensions and vibration to eliminate entrapped air.
The second – BS EN 12390-1:2012 is the UK implementation of the European
standard for testing hardened concrete under compressive loading. We followed the
selected standards to ensure that our final product can be directly applied on any
construction site within the European Union.
Figure S8 shows all the steps used for the preparation of concrete cubes:
mixing sand, cement, aggregates and graphene dispersed in water (shown in Fig
S8a) using a standard concrete mixer (Fig S8b), followed by the preparation of
cubes by pouring the concrete paste into moulds (Fig S8c), curing the cubes in water
(Fig S8d) and testing (Fig S8f).

Figure S8. Preparation and testing of concrete samples.

The control group with tap water concrete cubes was compared to 5 different
graphene reinforced ones, using the concentrations from the table S1. There is one
additional experimental group in which there was no graphite in the solution but
sodium cholate only.
 Fig. S9. A) The evolution of Young’s modulus of the material (Ec) over time (t(days)) comparing
the control group Tap Water (TW) to all reinforced experimental groups, including the surfactant-
only; Similarly, b) is the evolution of concrete compressive strength (fc in Mpa), over time
(t(days)).

S8. Concrete reinforcement with graphene suspensions produced

by sonication

Figure S10 includes data obtained from testing concrete reinforced with graphene
suspensions produced via sonication of FG solutions and blending large quantities of
IG solutions (50mg/ml). In these samples, the water-to-cement ratio was 0.4 and the
samples were hand tampered, which resulted, as expected, in overall lower fc
values. This was due to decreased workability of the fresh mix and many more air
bubbles entrapped within the matrix. Nevertheless, this study showed that the
reinforced concrete behaves in the same way – with highly increased early age
strength of all experimental groups and then continuous and more gradual increase
with time.
Fig. S10. A) Experimental study comparing the maximum compressive strength as a function of
time of TW and FG reinforced concrete; b) similar to a) but this one including data from IG
reinforcement.

Therefore, we can conclude the use of large quantities is unsuitable for concrete
mixing due to a high amount of residual unexfoliated graphite, which is costly and
difficult to remove. Even though the sediment was decanted prior to mixing, there
were still quite a lot of bulk particles within the liquid and this had a reverse
(negative) effect on the concrete hydration reaction. Fig. S10 shows experimental
data for fc of both TW and reinforced concrete. The different solutions on Fig. S10a
were prepared using the sonication method (labelled as 48h Sonication) while FG
was blended using the high-shear mixing. On Fig. S10b all FG and IG are prepared
using the high-shear blending. To further discuss the two methods of suspensions
preparing, photographs of the solutions were taken and plotted on Fig. S11. It can be
clearly seen that high-shear blending disperses the initial graphite powder
concentration much better than sonication method.
 Fig. S11. A) FG Solution produced by 48h of sonication. The sediment of unexfoliated graphite can
be identified at the bottom of the beaker; b) 4 different FG solutions all mixed by high-shear exfoliation
showing uniform dispersion of initial graphite powder.

S9. Calculating flexural modulus of elasticity from beam theory

principles

The flexural modulus of elasticity (Ecr) depends on the slope of the tangent line of the
first half of the stress-strain curve extracted from the 3-point bending test. From
basic beam theory the mid-span defection (0) is given by:

(1)

where L is the distance between the two outer supports, P is the load applied at the
middle of the beam, and I is second moment of area defined by:

(2)

with a and b the cross-sectional dimensions of the beam. Ecr can then be readily
calculated using the slope of the stress-strain curve:

(3)

From eq. (3) we calculated the flexural modulus of elasticity and plotted it as a
function of increasing graphene concentration and comparing it to the standard
concrete’s one.

S10. Prediction of total amount of cement used using graphene as

reinforcement for industrial purposes

As it is well-known, the cement production accounts for around 7% of the global

carbon emissions. The addition of graphene as reinforcement increases both the
compressive and flexural strengths of concrete which means that less amount of
concrete material can be used while still fulfilling the required specifications for the
loading of buildings. This relationship is shown with a simple equation:
Thus with the inclusion of graphene in nowadays concrete we could reduce the
amount of cement required for the original strength by nearly 50%.
The production of 1m3 of concrete requires ~ 360kg of cement (assuming 1:1.5:3
materials ratio, 0.45 w/c). Therefore the addition of 125g of graphene ($0.45 per
gram [4]) can decrease the total volume of cement down to 148kg per 1m 3. From
environmental point of view those numbers can account for a total decrease of 446
kg/tonne emissions of CO2, which is >300% greater than the overall reduction from
1990 until 2010 alone [5].

S11. X-Ray Diffraction Study of Graphene-Cement composite

We performed XRD study of the hydration crystals of cement, comparing Standard

Cement to FG and IG reinforced one, left to hydrate and harden for 24h. We identify
the major crystalline groups, as suggested in [6], and indicated the major differences
in peak ratios for the 3 experimental groups, focusing on different regions of the
spectra. Looking at the XRD spectra above, we can clearly identify sharper peaks at
23deg, 25.5deg, 38.5deg, 39.5 deg, 41deg and 43.5deg due to the incorporation of
graphene. This clearly suggests nanomodification of the cement crystals, more
specifically to Calcium Alluminoferrite, Calcium Carbonate, Tri- and Di- calcium
silicates as well as Calcium Aluminates groups. This suggests that FG and IG
graphene dispersions are interacting with various groups of the hydration crystals
family (listed in [7]) which suggests uniform distribution during mixing and hardening
stages. We believe that these microstructural changes in the crystals at the very
early stages of hydration are responsible for the ultra-high strength of the concrete at
later stages (7 and 28 days).
References
[1] K. R. Paton et al., Nature Materials 2014, 13, 624-630.
[2] E. Varrla et al., Nanoscale, 2014, 6, 11810
[3] F Withers, et al, Nano letters 2014, 14, 3987-3992.
[4] www.cheaptubes.com/product/graphene-nanoplatelets/
[5] WBCSD. CSI Progress report, June 2012
[6] D. Johnson, B. P. Dobson and K.S. Coleman, Current Opinion in Colloid &
Interface Science 2015, 20, 367-382.
[7] Paul E. Stutzman, Pan Feng and Jeffrey W. Bullard, Journal of Research of the
National Institute of Standards and Technology 2016, 121.
[8] I.G. Richardson, Cement and Concrete Research, 2008, 38, 137 – 158.